Time flies, so obviously the reported uptime may differ a bit (as
already occasionally noticed on CI); we cater to that by allowing a
delay of up to one second.
This reverts commit edccf32f7f.
This was reported to cause issues for as yet unknown reasons in
bug #78769. As this was intended as code cleanup, revert this from
7.4 at least. May reapply it to master later.
The string held by the zend_type may be released if the property
type gets resolved to a CE. I initially wanted to fix this by
storing a zend_type* instead (so the property type resolution
propagates to the ReflectionType), but decided against this in
light of upcoming union types support, where we also need to
represent parts of the union, and will not have a single zend_type*
we can reference.
These were checking whether the instruction set is supported by
the host CPU, however they were only used to condition on whether
this instruction set is targeted at all. It would still use dynamic
dispatch (e.g. based on ifunc resolvers) to select the actual
implementation. Whether the target is guaranteed to support the
instruction set without dispatch is determined based on pre-defined
macros like __SSE2__.
This removes the configure-time builtin cpu checks to remove
confusion. Additionally this allows targeting an architecture that
is newer than the host architecture.
First, the limitation already doesn't trigger if you copy the whole
file (i.e. use copy() or stream_copy_to_stream() and don't specify
a length). This happens because length will be 0 at the time of the
check and only later calculated based on the file size. This means
that we're already completely blowing the length limit for what is
likely the most common case, and it doesn't seem like anyone complained
about that.
Second, the premise of the code comment ("to avoid runaway swapping")
seems incorrect to me. Because this performs a file-backed non-private
mmap, no swap backing is needed for the mapping. Concerns over "memory
usage" are also misplaced, as this is a virtual mapping.
Use the shared empty array from ZVAL_EMPTY_ARRAY
For code that created an 10 arrays of 100000 empty arrays
(has the same result with `$assoc=true` and `{}`)
- This is the worst-case comparison, but I'd expect 0-length arrays to be fairly
common in regular data for json_decode
- The parser implementation was using function pointers so that third party
extension developers could reuse the json parser for their own
data structures, etc. (I think).
This PR is meant to let those third party extensions continue working
without changes.
Before this patch: In 0.126 seconds: added 97.99 MiB
After this patch: In 0.096 seconds: added 41.99 MiB
```php
<?php
$json = '[' . str_repeat('[],', 100000) . "null]";
$start_memory = memory_get_usage();
$start_time = microtime(true);
$result = [];
for ($i = 0; $i < 10; $i++) {
$result[] = json_decode($json);
}
$end_memory = memory_get_usage();
$end_time = microtime(true);
// Before this patch: In 0.126 seconds: added 97.99 MiB
// After this patch: In 0.096 seconds: added 41.99 MiB
printf("In %.3f seconds: added %.2f MiB\n", $end_time - $start_time, ($end_memory - $start_memory)/1000000);
// For objects
$json = '[' . str_repeat('{},', 100000) . "null]";
$start_memory = memory_get_usage();
$start_time = microtime(true);
for ($i = 0; $i < 10; $i++) {
$result[] = json_decode($json, true);
}
$end_memory = memory_get_usage();
$end_time = microtime(true);
// Before this patch: In 0.126 seconds: added 97.99 MiB
// After this patch: In 0.096 seconds: added 41.99 MiB
printf("In %.3f seconds: added %.2f MiB (objects decoded as arrays) \n", $end_time - $start_time, ($end_memory - $start_memory)/1000000);
```
Closes GH-4861.
To work around the limitation of the current rudimentary vectorcall
support in our patched libffi, we forbid yet unsupported declarations,
i.e. float/double parameters at certain positions (SIMD vector types
and HVA types are not supported anyway).
For some reason FD 120 seems to exist on macos quite often, while
FD 12 did not... Let's try an even larger number, otherwise we
should just drop this test.
We need to install a number of additional packages that are installed
by default on the AMD64 workers.
We also have to manually set up the MySQL user.
For now we don't set up Postgres -- if anyone wants to figure that
out, it would be great ;)
Log redirections in compile.sh are removed, because /dev/stdout is
not accessible. We don't see to use this anyway.
Assigning to an uninitialized typed property will no longer trigger
a call to __set(). However, calls to __set() are still triggered if
the property is explicitly unset().
This gives us both the behavior people generally expect, and still
allows ORMs to do lazy initialization by unsetting properties.
For PHP 8, we should fine a way to forbid unsetting of declared
properties entirely, and provide a different way to achieve lazy
initialization.
instanceof_function() requires linked classes. I'm not reusing
unlinked_instanceof() here, because it performs class loading,
which wouldn't be right here, I think.
Parent interfaces are copied into the interface list during
inheritance, so there's no need to perform a recursive check.
Only exception are instanceof checks performed during inheritance
itself. However, we already have unlinked_instanceof for this
purpose, it just needs to be taught to handle this case.
Closes GH-4857.
instanceof_class does not need to check for a NULL pointer in the
first iteration -- passing NULL to this function is illegal.
instanceof_interface does not need to use instanceof_class(), it
only has to check whether the CEs match exactly. There is no way
for an interface to appear inside "parent", it will always be in
"interfaces" only.
The instanceof_interface_only() function was dead code (always
returned zero).
Clarify that the last parameter indicates whether the passed CE
is interface or class and rewrite the code in terms of assertions.
I'm not totally sure, but I have a strong suspicion that the fact
that this produces an error is an artifact of undefined cast behavior
(which will yield INDVAL on x86 but saturate on ARM). INF seems to
be the only value that results in an error even on x86 (variations
like -INF or NAN succeed).
It might make sense to just remove this test entirely, but for now
let's skip it on non-x86.
This stat property seems to be somewhat unreliable depending on the
filesystem. On Travis ARM64 CI a much larger payload is required
to get this value to increase.
Noticed while working on union types: We do not load argument and
return types during type checks, but we do load property types.
I'm normalizing the behavior towards the existing status quo (not
loading), though we may consider loading everywhere (all types,
and instanceof) in order to properly support class aliases.
Fix PHP-FPM failure in the case of concurrent reload attempts.
Postpone signal delivery to the fpm master process till proper signal
handlers are set. Prevent the following case:
- Running master process receives `SIGUSR2` and performs `execvp()`.
- Another `SIGUSR2` is arrived before signal handlers are set.
- Master process dies.
- Requests to the HTTP server handled by PHP-fpm can not be served
any more.
Block some signals using `sigprocmask()` before `execvp()` and early
in the `main()` function. Unblock signals as soon as proper
handlers are set.
Fixes bug #74083
Emitting errors is fairly expensive, to the point that parsing
a file with a huge number of invalid tags can take seconds.
Generating ten thousand errors is unlikely to help anybody, but
constitutes a potential DOS vector.
Update array access syntax deprecated in line 175 and 204
Deprecated: Array and string offset access syntax with curly braces is deprecated in /home/alex/php/hello/ext_skel.php on line 175
Deprecated: Array and string offset access syntax with curly braces is deprecated in /home/alex/php/hello/ext_skel.php on line 204
Property types are invariant, but may still have to load classes in
order to check for class aliases. This class loading should follow
the same rules as all other variance checks, rather than just
loading unconditionally.
This change integrates property type invariance checks into the
variance system as a new obligation type, and prevent early binding
if the type check cannot be performed.
This goes in the reverse direction of 4463acb951.
After looking around a bit, it seems that we already check for
Z_ISERROR_P() on the get_property_ptr_ptr return value in other places.
So do this in zend_fetch_property_address() as well, and also make
sure that EG(error_zval) is indeed returned on exception in
get_property_ptr_ptr.
In particular, this fixes the duplicate exceptions that we used to
get because first get_property_ptr_ptr threw one and then
read_property throws the same exception again.
Relying on setting ERROR if an exception happened during the
property address fetch is both a bit fragile and may pessimize
other codepaths that will check for exceptions in the VM. Adding
an extra exception check instead, which should also allow us to
drop the use of ERROR in this area in master.
As of PHP 7.4.0, the `get_property_ptr_ptr` handler is mandatory; we
implement it to always return `NULL`, which is equivalent to not
setting the handler in former versions.
We add a portable and faster test case than what has been presented in
the bug ticket.
After fixing the int->double coercion case, this is already verified
at compile-time, so there is no need to redo this type check on
every call.
Only perform the type check every time for the case of AST default
values.
Previously if the "non well formed" notice was converted into an
exception we'd still end up executing the function.
Also drop the now unnecessary EG(exception) checks in the engine.
Additionally remote a bogus exception in zend_is_callable: It
should only be writing to error, but not directly throwing.
Make sure the initialize the result of FETCH_OBJ_UNSET operations.
I'm using a NULL value rather than ERROR here, because the latter
no longer exists in master.
Usually it will already fail when opening, but reads can also
fail since PHP 7.4, in which case we still need to place the
file handle in open_files to make sure the destructor will run
on it.
* pack() only requires one argument
* stream_context_set_option() only requires two arguments
* ReflectionMethod::getClosure() accepts no args for static methods
* DOMDocument::createProcessingInstruction() only requires one arg
* DOMImplementation::createDocument() only requires two arguments
* DOMDocument::importNode() only requires one arg
* mysql_get_client_version() doesn't accept any args,
despite what the docs say...
Otherwise we would try to access an array element of `false`, which
issues a notice as of PHP 7.4.0. This would happen, for instance, for
bug63447_001.phpt if CGI is not available.
The default depth is large enough to cause stack overflows in
msan builds, though apparently only on old clang versions. Avoiding
a stack overflow there requires making the depth *much* smaller,
less than 1000. As I don't think that's reasonable for all the other
setups where 4k works fine, I'm just dropping this part of the test.
This test is somewhat fragile in that it depends on how well a
particular regex is optimized. Apparently on 6.9.1 this regex
would hit the default retry_limit of 1000000 already. I'm limiting
this to 6.9.3 because that's the version that works for me.
This is very similar to the existing mbstring.regex_stack_limit,
but for backtracking. The default value matches pcre.backtrack_limit.
Only used on libonig >= 2.8.0.
I'm going for a very conservative fix here, where the previous
logic is restored for the case where an object is passed to
method_exists(). We might want to check against EG(scope) instead,
but this seems like a safer choice.
This means that behavior in PHP 7.4 changes only for
method_exists('C', 'privateMethodNotOnC'), which should be sensible.
Not NULLing the static_variables pointer for shadow methods during
static var shutdown would be a way to avoid this leak, but unless
there's evidence that inherited private methods with static vars are
actually a common use-case, I don't think we should keep this kind
of fragile edge-case optimization.
Fixes OSS-Fuzz #17875.
Add a max_depth option to unserialize and an unserialize_max_depth
ini setting, which can be used to control the depth limit. The
default value is 4096.
This option is intended to prevent stack overflows during the
unserialization of deeply nested structures.
This fixes bug #78549 and addresses oss-fuzz #17581, #17589, #17664,
and #17788.
Around a quarter of all strings in array tokens would have a string that's one
character long (e.g. ` `, `\`, `1`)
For parsing a large number of php files,
The memory increase dropped from 378374248 to 369535688 (2.5%)
Closes GH-4753.
rename_variation2.tmp is ... unsurprisingly ... already used by
rename_variation2.phpt. Alternatively it could be placed in the
directory that the test creates, but maybe this is intended to
test cross-directory renames, so I'll leave it.
Resources used as array keys are generally handled by throwing a
notice and converting the resource to the resource handle. The only
exception is the [$resource => null] syntax, where this was treated
as an illegal offset type instead. However, this also only happened
for VM evaluations, the AST evaluator did handle resources correctly.
Code coverage reports (`make lcov`), since commit eef8522 (7.4 branch),
generates incorrect coverage and emits warnings. Simplifying the
Makefile.gcov file has the side-effect of resolving the issue.
Processing sapi/cli/php_http_parser.gcda
php-src/lcov_data/sapi/cli/php_http_parser.gcda:stamp mismatch with notes file
geninfo: WARNING: gcov did not create any files for php-src/lcov_data/sapi/cli/php_http_parser.gcda!
...
Processing ext/mbstring/mb_gpc.gcda
php-src/lcov_data/ext/mbstring/mb_gpc.gcda:stamp mismatch with notes file
geninfo: WARNING: gcov did not create any files for php-src/lcov_data/ext/mbstring/mb_gpc.gcda!
Closes: https://bugs.php.net/bug.php?id=52718.
See also: https://bugs.php.net/bug.php?id=78288.
Don't use <value> type for JSON tokens that don't have a value
and remove the errlex productions -- we're going to get an
unexpected token error anyway, there's no need to handle these
explicitly.
This also removes the awkward workarounds for the unused value
warnings.
libsodium measures the memory cost in bytes, while password_hash() and
friends expect kibibyte values. We have to properly map between these
scales not only when calling libsodium functions, but also when
checking for allowed values.
We also refactor to rid the code duplication.
To clean up the mess here a bit, check for invalid octal digits
with an explicit loop instead of mixing this into the string to
number conversion.
Also clean up some type usage.
The ZEND_ALLOCATOR() macro compares against constants that are
larger than the unsigned short type, resulting in warnings on
clang. Avoid this by explicitly casting to size_t.
This makes sure that we use compiler builtins when they are
available and thus avoid ubsan warnings in clang.
And also reindent the fallback implementation.
zend_emit_op_data may reallocate the op_array, so the assignment
of the RETURNS_FUNCTION flag may happen on an outdated opline.
Restructure the code a bit to set the flag before calling
zend_emit_op_data().
This is a fix for symfony/symfony#32995.
The behavior is:
* Throwing exception when loading parent/interface is allowed
(and we will also throw one if the class is simply not found).
* If this happens, the bucket key for the class is reset, so
it's possibly to try registering the same class again.
* However, if the class has already been used due to a variance
obligation, the exception is upgraded to a fatal error, as we
cannot safely unregister the class stub anymore.
The requirements for parent/interface are difference than for the
variance checks in type declarations. The latter can work on fully
unlinked classes, but the former need inheritance to be essentially
finished, only variance checks may still be outstanding.
Adding a new flag for this because we have lots of space, but we
could also represent these "inheritance states" more compactly in
the future.
Perform PHPUnit installation under php7.3 -- which will also make
it work on master.
Also properly resolve the tty issue by specifying excluded groups
during the test run.
microtime() doesn't return an array,
and gettimeofday() doesn't return a string.
See _php_gettimeofday in microtime.c (mode is non-zero for gettimeofday)
Most bundled libraries in PHP code already include license information
in the dedicated README.REDIST.BINS file. Instead of complicating
documentation files more this patch only updates the PCRE license info
in the readme file.
The newest version we're checking (libssldap50) seems to be about
15 years out of date. We could add support for libssldap60 (also
unmainted, but more recent), but given how nobody has expressed any
interest in this, I'm going ahead and dropping this code.
Split the default and custom path case. If the default is used,
assume that the library must be on the default include and lib path.
Only check that the version is appropriate.
Something similar is needed for ldap, but the checking code is much
more complex there, so I'm only adding a workaround for now.
We need to make sure that the function is fully compiled before we
calculate the stack size. There already was a check for directly
recursive calls, but the same issue exists with indirectly recursive
calls.
I'm using DONE_PASS_TWO as the indication that the function is
fully compiled.
There are two related changes here:
1. Also check for S_ISCHR/FILE_TYPE_CHAR when checking for pipes, so
that we detect ttys as well, which are also not seekable.
2. Always set position=-1 (i.e. ftell will return false) when a pipe
is detected. Previously position=0 was sometimes used, depending on
whether we're on Windows/Linux and whether the FD or FILE codepath
was used.
The exception mechanism assumes that exceptions from DO_FCALL are
already happening after the function call. This means that we are
currently leaking the passed arguments, and I think we can also
corrupt the VM stack due to incorrect frame linking in some cases
(there are assertion failures if the VM stack page size is reduced).
Instead handle the stack frame freeing manually for this special
case.
(excluding spl_autoload)
spl_object_id() is of the most interest to me,
since I frequently call it in an application.
This includes false/null types caused by wrong argument types and wrong argument
counts.
I can't rule out iterator_to_array returning null in spl_iterator_apply,
so leave MAY_BE_NULL in.
With review comments by nikic:
Co-Authored-By: Nikita Popov <nikita.ppv@googlemail.com>
Run some open-source projects through an aggressive debug
configuration with asan and ubsan. We don't care about test results,
only check that we don't assert or crash.
Currently testing laravel, symfony and amp.
When building PHP outside of the source tree:
mkdir custom-build-dir
cd custom-build-dir
../path/to/php-src/configure
The directories need to be manually created including the pear directory
so the pear installation PHAR file doesn't need to be downloaded from
the remote location.
Closes GH-4639
While we don't need to give this method separate static vars, we
do still need to perform an addref, as there will be a corresponding
delref in the dtor.
These flags have been deprecated in glibc 2.28, so we also
deprecate them in PHP.
As we can't deprecate constants, we can only check for their use
in socket_addrinfo_lookup().
Only deprecate unbinding of $this from a closure if $this is
syntactically used within the closure.
This is desired to support Laravel's macro system, see laravel/framework#29482.
This should still allow us to implement the performance improvements
we're interested in for PHP 8, without breaking existing use-cases.
Also remove the hack where scope is set to NULL in order to make
free_internal_arg_info work. Instead explicitly call it for class
methods.
This fixes the asan build for Zend/tests/bug77494.phpt.
This allows us to drop the intermediate allocation for
spl_pqueue_elem.
This fixes GC for SplPriorityQueue, because we can now directly
return a well-formed GC child buffer.
When storing two temporary variables for delayed __unserialize() calls,
we have to make sure that both fit into the same linked list element.
To that end we introduce the internal API `tmp_var` which allows to
reserve `num` slots in the same list element.
We also fix the `var_dtor_entries` struct definition to use the proper
size, namely `VAR_DTOR_ENTRIES_MAX`.
We must not assume that any file which is not a directory is a regular
file. Therefore we employ `GetFileType()` in this case to properly
distinguish between character special, FIFO special and regular files.
This removes an incorrect optimization (I think this code used to be
necessary to properly handle references in the Serializable based
implementation, but now this code just avoids an array duplication
in a way that is not sound).
For objects with destructors, we will now only call the destructor
in the initial GC run, and remove any nested data. The object is
marked purple so it will be considered a root for the next GC run,
at which point it will be fully destroyed, if possible.
GC counts change on a number of tests, as the objects now get
destroyed later.
Checking for the exact linker version appears to be too restrictive; it
should be fine if the tens match.
We also refactor to avoid repeating ourselves.
The fdatasync check has been removed since PHP 5.3.
The Solaris fix was introduced via
8d63360fc4
and is today no longer relevant since the sqlite library is neither
bundled in php-src anymore neither the check needs to be done via the
PHP build system.
Closes GH-4485
Make sure that fpm_event_add calls inside a timer callback work by
unregistering the event from the queue before invoking its callback.
The read timeout in tester.inc is increased because the added test
needs two seconds (one for SIGTERM, one for SIGKILL) until the
reload succeeds, so we should wait longer than that for a response.
Argument unpacking may need to create references inside the array
that is being unpacked. However, it currently can only do this
if a plain variable is unpacked, not for any nested accesses,
because the value is fetched for read. Resolve this by fetching
the operands for RW.
When performing a constant visibility check during compilation we
might be dealing with unlinked classes and as such should account
for the possibility of unresolved parents.
Even if we know the file size, we still need to read in a loop in
case the read call returns an incomplete result.
This was less of an issue previously because we did not use the
"one large read" approach for non-plain stream wrappers.
These two files have been: "UTF-8 Unicode (with BOM) text".
By applying `dos2unix` on these files the BOM has been removed.
I checked the whole source code with dos2unix:
These were the only two text files affected.
This was added in 7.1 when add_assoc_string mistakenly accepted
a char* rather than const char* parameter and is no longer needed.
We can use SSL_CIPHER_get_version() directly.
Make sure we deref the OBJ_IS result, because we store it in a TMP
var, which is not allowed to contain references and will cause
assertion failures in the unspecialized VM.
This also partially reverts fd463a9a60,
which merged the TMP and VAR specializations of COALESCE to work
around this bug.
An alternative would be to change the result type of OBJ_IS back
to VAR.
By avoiding unused variable opline warnings. Also clean up the
replacement of ZEND_VM_SPEC -- we were sometimes treating it as
an always-defined constant with a value (what it actually is) and
sometimes as a conditionally defined constant (which it isn't, but
which still worked thanks to the specializer). Switch to only
treating it as a constant with a value.
Symbols are not used in the current code.
Also remove not needed ocijdbc8 library addition and checks whcih were
removed also via b8e1d4d4e5 so Oracle 8
is no longer supported.
Closes GH-4454
The php_stream_read() and php_stream_write() functions now return
an ssize_t value, with negative results indicating failure. Functions
like fread() and fwrite() will return false in that case.
As a special case, EWOULDBLOCK and EAGAIN on non-blocking streams
should not be regarded as error conditions, and be reported as
successful zero-length reads/writes instead. The handling of EINTR
remains unclear and is internally inconsistent (e.g. some code-paths
will automatically retry on EINTR, while some won't).
I'm landing this now to make sure the stream wrapper ops API changes
make it into 7.4 -- however, if the user-facing changes turn out to
be problematic we have the option of clamping negative returns to
zero in php_stream_read() and php_stream_write() to restore the
old behavior in a relatively non-intrusive manner.
First step when creating the `configure` script is currently using
make. This is helpful when developing buildsystem to only rebuild
`configure` and `main/php_config.h.in` files when one of the *.m4
or configure.ac file changes and saves the developer time a little.
Realistically however, it is not needed that much considering the
times of several seconds to fully rebuild the configure script and
configuration header. The next step when running the `configure`
script is much more time consuming so performance on buildconf
level is a bit questionable and won't be noticed on today's
systems.
Additionally:
- buildconf now removes cache and all targets and uses -f option on
the first step i.e. autoconf. The autoheader does not need the -f
option in this case.
Closes GH-4437
- HAVE_NET_SNMP removed via cab643f615
- HAVE_CRYPTO_AEAD_AES256GCM ad120c5ae9
- Remove HAVE_FREETDS and HAVE_LIBDNET_STUB (not used in current
extension; copy paste from other removed extensions)
Closes GH-4448
- when $status is boolean, E_NOTICE appears and tests fail
- opcache is never enabled on these tests anyway.
- Add opcache.revalidate_freq=0 to testing script
Instead of checking this in the tests files it can be used directly in
the main testing script. Also this fixes failed tests.
- Use opcache_invalidate() instead of sleep() delay in some tests. Some
delays are still used since the opcache_invalidate seems to be buggy
and not working as expected.
Closes GH-4392
Oniguruma is exclusively used by ext/mbstring, and only if mbregex is
enabled. Therefore it is unnecessary and confusing to have Oniguruma
related config stuff scattered elsewhere.
While we're at it, we also remove the referral to the bundled libonig
which is removed as of PHP 7.4.0, and the duplicated call to
`PHP_INSTALL_HEADERS()`.
The classic Windows console only supports the Unicode BMP, so it can't
handle the LIGHTNING MOOD character (U+1F5F2). It also seems that
common fonts do not support the LIGHTNING (U+2607) character, so a
general removal is appropriate.
However, Windows is already "ready for the future"[1], but PHP for
Windows is not (yet).
[1] <https://devblogs.microsoft.com/commandline/windows-command-line-unicode-and-utf-8-output-text-buffer/>
ext/win32std is unmaintained for years, so it is highly unlikely that
it is available. We therefore use regedit instead to register the
event source. We also remove the bundled syslog.reg, since it is of no
use, unless generated for the respective installation.
Used in php-src the past and today removed and not used anymore:
- HAVE_CURL_EASY_STRERROR
- HAVE_CURL_MULTI_STRERROR
- HAVE_NEW_MIME2TEXT
- HAVE_MBSTR_CN
- HAVE_MBSTR_JA
- HAVE_MBSTR_KR
- HAVE_MBSTR_RU
- HAVE_MBSTR_TW
Part of oniguruma which doesn't use these anymore
- NOT_RUBY
- HAVE_STDARG_PROTOTYPES
Unused:
- HAVE_MPIR
Closes GH-4427
Formerly, there was at most a single OPcache instance per user and the
so called system ID (which is determined from the PHP version).
Sometimes multiple OPcaches might be desired, though, particularly for
unrelated CLI scripts, which may even be necessary (e.g. for our test
suite in parallel mode).
We therefore introduce a new INI directive `opcache.cache_id` which
allows to configure independent OPcache instances for the same user.
We also use `GetUserNameW()` instead of `php_win32_get_username()`,
because the latter retrieves the user name encoded in the
`default_charset`, which can obviously yield different results for
different charsets, leading to OPcache "incompatibilities". Slightly
worse, some characters may not even be encodeable in the
`default_charset` and would be replaced by question marks, which could
result in different users sharing the same OPcache.
We also refactor, and re-use existing APIs to avoid duplicated code.
This reverts commit cc44bad1f6, since its
assumptions were not correct. Actually, the classic event logging is
still used by PHP, because Windows 7 is still to be supported for a
while. Without the respective registry entries, the event log entries
show an error message regarding missing ID descriptions.
Thanks to ab@php.net for hinting at this!
Obviously, the part depending on ext/win32std has still to be
overhauled, and further improvements are conceivable; we will address
this in due course.
By adding a flag to avoid forced fstat for includes. The two fstats
will happen back to back and we don't care about a possible
invalidation.
I was hoping to move this higher up in the stack and make the
ISREG check somewhere in fsizer of fixup, but this doesn't really
seem to be possible. E.g. an FP stdin handle will not be a regular
file but of course needs to be allowed. Additionally custom stream
wrappers may not implement this functionality.
If we're including a file via PHP streams, we're not going to trust
the reported file size anyway and populate in a loop -- so don't
bother determining the file size in the first place. Only do this
for non-tty HANDLE_FP now, which is the only case where this
information was used.
Disable buffering in PHP streams, to avoid storing and copying the
file contents twice.
This will call stream_set_option() on custom stream wrapper as
well, so the method needs to be implemented to avoid a warning.
- yp_get_default_domain was part of ext/yp
- functions checks produce HAVE_function_name symbols. These checks are
currently not used in php-src neither in the extensions out there.
- Removed symbols because they are not used in the code:
- HAVE_GCVT
- HAVE_PUTENV
- HAVE_PUTENV
- HAVE_SETVBUF
- HAVE_TEMPNAM
- HAVE_SIN (sin is also defined in C89 standard)
- HAVE_SETSOCKOPT
- HAVE_LOCKF
- HAVE_ISASCII
- HAVE_YP_GET_DEFAULT_DOMAIN (and other yp extension related unused checks)
- HAVE_LINK
- HAVE_USLEEP is already defined in Windows configuration header
- HAVE_LIBBIND has not been used in php-src for a while anymore
- HAVE_GETHOSTNAME is duplicated in Windows configuration header
Closes GH-4417
Using mmap() is unsafe under concurrent modification. If the file
is truncated, access past the end of the file may occur, which will
generate a SIGBUS error. Even if the length does not change, the
contents may, which is a situation that the lexer certainly is not
prepared to deal with either.
Reproduce with test.php:
<?php
file_put_contents(__DIR__ . '/test.tpl',
'AAA<?php $string = "' .
str_repeat('A', mt_rand(1, 256 * 1024)) .
'"; ?>BBB' . "\r\n");
require_once __DIR__ . '/test.tpl';
And:
for ((n=0;n<100;n++)); do sapi/cli/php test.php & done
Instead of handling shebang lines by adjusting the file pointer in
individual SAPIs, move the handling into the lexer, where this is
both a lot simpler and more robust. Whether the shebang should be
skipped is controlled by CG(skip_shebang) -- we might want to do
that in more cases.
This fixed bugs #60677 and #78066.
Since mysqlnd is solely developed in the php-src repository, it makes
not much sense to have its own version number, particularly since
nobody cares to update it. Therefore we use PHP's version number, but
stick with the "mysqlnd" prefix which can be important to distinguish
the mysqli driver (libmysqlclient vs. mysqlnd).
On this benchmark:
function simple_string_escape() {
$a = "test'asd'asd'' asd\'\"asdfasdf";
for($i=0; $i<512; $i++) {
$a .= chr($i%256);
}
for ($i = 0; $i < 100000; $i++) {
if ($a === stripslashes(addslashes($a)))
$a .= chr($i%256);
else {
echo "error at i=".$i."\n";
return;
}
}
}
the execution time goes from 21.619s to 8.139s (165% speedup) on an A1 Graviton instance.
When removing the characters that need escaping, i.e., this benchmark:
function simple_string() {
$a = "testasdasd asdasdfasdf";
for ($i = 0; $i < 10000; $i++) {
if ($a === stripslashes(addslashes($a)))
$a .= "test dedeasdf";
else {
echo "error at i=".$i."\n";
return;
}
}
}
the execution time goes from 2.932s down to 0.516s (468% speedup) on an A1 Graviton instance.
The undef PTHREADS converts to define if thread safety is configured.
This step is already done by pthreads m4 macros from TSRM so this now
removes duplicated PTHREADS defines from php_config.h.
This test does not work on Windows, if the child process has OPcache
enabled via php.ini, since TEMP is not set in the environment, so
OPcache can't find the base address file, and terminates the childs.
To avoid this (and potentially some other issues regarding missing
environment variables, such as PATH), we force the child process to
ignore any ini files.
The fix for bug #78241 assumed that `time_t` would always be 64bit, but
actually is 32bit for x86. We therefore enforce 64bit arithmetic to
avoid wrapping.
- Checks and defines are not relevant for files that include it anymore
- Some code is not used anymore
- Defines are a bit duplicated in zend_portability.h and TSRM.h file
- MAXPATHLEN defs moved to zend_virtual_cwd.h
This adds support for doing something like:
[1 => ['pipe', 'w'], 2 => ['redirect', 1]]
This will make descriptor 2 on the child end a dup'd descriptor 1.
This is mainly useful in conjunction with shell-less mode, because
we don't have an easy way to do "2>&1" there.
Additionally we support:
[1 => ['pipe', 'w'], 2 => ['null']]
Which would be the same as a >/dev/null or >nul redirect, depending
on platform.
export() methods were implemented in a roundabout way, where they
would call Reflection::export(), which would then call __toString().
Cut out the middleman by directly calling __toString().
The test currently assumes that we'll first read the data of
stdout and stderr and then see eof on stdout and stderr. However
we could also read stdout, see eof on stdout, read stderr and see
eof on stderr, depending on timing.
Avoid output ordering issues by collecting events into a per-pipe
array, so interleaving is not visible.
The time goes from 0.838s down to 0.029s (a 28x speedup) on a Graviton A1
instance and the following benchmark:
function simple_crc32() {
$a = "foo";
for ($i = 0; $i < 10000; $i++) {
crc32($a);
$a .= "o".$i;
}
}
Instead of patching configuration headers template generated by
the given tools - autoheader, this moves patching these symbols to
the configure step before creating and invoking the config.status
and before the configuration header file is generated from the
patched template.
Closes GH-4374
No notice is thrown for list() accesses, because we did not come
to an agreement regarding patterns like
while ([$key, $value] = yield $it->next()) { ... }
where silent null access may be desirable.
No effort is made to suppress multiple notices in access chains
likes $x[0][0][0], because the technical complexity this causes
does not seem worthwhile.
RFC: https://wiki.php.net/rfc/notice-for-non-valid-array-container
- Git is used for moving existing extensions to PECL
- PHP versions with the PHP 7.4 examples
- Some minor tweaks here and there
- Mention re2c also
- Mention Docker image as a helper tool
And don't return null for rc=1 references. Leave it to the user
to decide whether or not they want to consider these as references
or not.
Fixes bug #78263.
Changes:
- Coding style fixes
- ${1+"$@"} replaced with "$@" [1]
- EXTRA_MODULE_PTRS variable is not used anymore
- awk tool defined via environment variable
- srcdir determined automatically from the genif.sh location
[1] https://www.in-ulm.de/~mascheck/various/bourne_args/
Closes GH-4372
Another stab in the dark to fix the intermittent failures of timeout
tests on macos CI: We're using ITIMER_PROF, which means that the
timer counts against user+system time. The "busy" wait loop counts
against real time. Currently it calls microtime() on every iteration.
If that call is implemented as a syscall rather than going through
vDSO or commpage we might be seeing many context switches here which
drive up the real time, but not user or system time.
See if making the loop busier and calling microtime() less helps the
situation.
These are part of the C89 and on today's systems not needed to be
checked anymore. This removes symbols HAVE_SIGNAL and HAVE_STRERROR.
- http://port70.net/~nsz/c/c89/c89-draft.html
- locale.h is also part of C89 but will be removed per request in PHP 8
This simplifies the libmagic patch:
- in upstream the HAVE_STRERROR check has been removed
- in upstream library the HAVE_SIGNAL_H has been removed
- indentations syncs with the upstream library
- some irrelevant changes removed from the patch (log comment), upstream
has this correctly logged already so no need to patch the comment.
Closes GH-4298
While it is already possible to *set* CFLAGS and LDFLAGS (actually all
variables) from the environment for `nmake` (by passing the `/E`
option), it is not possible to *add* any (C|LD)FLAGS, which can be
useful in some cases. Instead of allowing this for `nmake`, we add
support for additional custom (C|LD)FLAGS to `configure`, similar to
how that works on Linux, so one could actually write:
````
set CFLAGS=foo & set LDFLAGS=bar & configure
````
This also allows us to use these flags during configure.
Although TsHashTable and the according API are supposed to easily make
a HashTable thread-safe, they do not; for instance, there can be race
conditions between finding and updating entries. We therefore avoid
the usage of TsHashTable in favor of a HashTable with our own mutex
management.
The patch has been provided by krakjoe@php.net; I only did some minor
fixes and tweaks.
- Variables php_abs_top_srcdir php_abs_top_builddir are no longer used.
- ZEND_EXT_TYPE is always zend_extension and variable is no longer used.
Closes GH-4378
This simplifies TSRM build steps a bit and avoids doing unnecessary
steps:
- The `PTHREADS_CHECK_COMPILE` can called inside the for loops only
since this is only where the `$pthreads_checked` variable is used.
- Assigning variables can be then done only in the configure.ac
once.
- use `m4_include()` instead of the `sinclude()` in the middle of
the build steps.
- The `$threads_result` variable is not used in the code or in
extensions.
This test hangs on Windows for ~10 minutes as of commit 30019f4, so we
temporarily skip it. The issue will be investigated by Joe and myself
in due course.
Conditional function declaration may be referenced by functions or
classes that have been preloaded, so we should not move them back
to scripts.
What we probably should be doing though is to discard conditional
functions that are not used in the optimizer. This is probably
reasonably common for polyfills, where we will be able to const-eval
the conditions and drop the BBs declaring the functions, but won't
delete the function declarations themselves.
LITERAL_VALUE with related literals cannot be merged, avoid inserting
them in the literals hash in the first place. Otherwise we may end up
inserting the same key multiple times. We may also miss merging
opportunities due to the duplicate keys.
Check whether there is a parent/interface/trait method with the same
name and only then require the type to be known. This reduces the
number of cases where this triggers in practice a lot.
The output of the three handles may be interleaved. We already have
curl_basic_018.phpt which uses curl_multi_getcontent() and thus
has predictable output. As such, I'm dropping this test altogether.
This adds the Autoconf's --enable-option-checking=fatal option so
when non existing option from the PHP's configure options is used
a fatal error happens.
Instead of building a custom macro for checking configure options,
Autoconf 2.62+ already outputs a warning at the beginning and the end
of the output of configure script. It automatically detects correct
and wrong options better.
This is related also to bug #55634.
So now instead a better way is the default Autoconf approach:
This outputs a warning at the beginning and end of the configure output:
./configure --with-non-existing
This results in fatal error:
./configure --non-existing
configure: error: unrecognized option: `--non-existing'
Try `./configure --help' for more information
The `--enable-option-checking=fatal` results in fatal error for all non
existing options:
./configure --with-non-existing --enable-option-checking=fatal
configure: error: unrecognized options: --with-non-existing
Closes GH-4348
This uses a separate job template, because msan requires all used
libraries (apart from glibc) to be instrumented, so we can't link
any external libraries unless we recompile them. As such, we need
a much more minimal configure.
This macro is not needed anymore. The AC_PROG_CC is done in the main
configure.ac file and the ranlib check is done by the bundled libtool
macros.
Closes GH-4339
The following functions don't need to be checked anymore since the
they are not used across the code or the symbols aren't used anymore:
- cuserid (not used)
- lrand48 (not used and removed via
6d6ef7aacc)
- random (check is not used)
- srand48 (not used)
- srandom (not used)
- strdup (check is not used)
and the unused check symbols:
- HAVE_CUSERID
- HAVE_LRAND48
- HAVE_RANDOM
- HAVE_SRAND48
- HAVE_SRANDOM
- HAVE_STRDUP
Closes GH-4338
The result of this check is to whether to define the HAVE_CMSGHDR symbol
or not. The HAVE_CMSGHDR is never used in the code and it has been
removed via 90289924c0.
Closes GH-4337
It would be great if this were fully initialized, but it's not
really a problem either (as long as we don't care about reproducible
file cache), so ignore this for now.
This has been sporadically failing on macos for a while. We've tried
a couple of things (using int arithmetic, normalizing to gettimeofday),
but this issue remains. I'm adding some tolerance to the test to
avoid CI failures.
Removed unused checks:
- mbsinit check removed, HAVE_MBSINIT removed (not used in php-src)
- mempcpy check removed, HAVE_MEMPCPY removed (not used in php-src anymore since
560ed89bfb which uses PHP's own implementation)
- strpncpy check removed, added via a8c9e893b6 and
not used.
- setpgid check removed since HAVE_SETPGID is not used
Moved to a central configure.ac:
- fpclass
- mbrlen moved to configure.ac (since the HAVE_MBRLEN is used accross the php-src)
- sigprocmask
- getcwd
- getwd
- glob
- strfmon
- nice
Duplicated checks removed:
- gethostname
- getlogin
- getpwuid_r
- socketpair
- mprotect check simplified
These were part of the following decomissioned extensions:
- PHP_SYBASE part of the sybase extension
- PHP_MSQL_* part of the msql extension
- PHP_APACHE_* symbols were used in apache and apache_hooks sapis
- PHP_FHTTPD_* symbols part of old sapi/fhttpd
- PHP_MYSQL_* symbols were part of the legacy mysql extension
- Remove PHP_ORACLE_* symbols
These were part of the obsolete oracle extension that was moved to PECL
via f35af61ef8.
- PHP_OCI8_SHARED_LIBADD and PHP_OCI8_DIR were removed via
c33f88ccb0
And other are not used anywhere across the php-src codebase
The limits.h header is part of the C89 and is today available
everywhere. There is no need to check for presence of this header
anymore.
The timelib has already been patched upstream via
aae5907cb7
PHP extensions out there shouldn't rely on symbols defined during the
build anyway and neither they do on this particular symbol anymore.
Just marking them as no_sanitize("memory") is unforunately not
sufficient, as the function still gets instrumented -- the attribute
only disables reporting.
Also adds an --asan flag to run-tests.php to setup all the necessary
environment variables. Some tests are marked as skipped because they
are incompatible with asan or too slow.
I'm basing this on the DEBUG_ZTS build, which seems to give us the
most mileage.
Available under -DZEND_TRACK_ARENA_ALLOC. This will use the system
allocator combined with arena checkpointing & release semantics
and allows analyzing arena usage under asan/valgrind.
I've sacrificed the duplicate arena implementation in mysqlnd, as
the integration with mysqlnd alloc is not worth the code duplication
to me.
In this case the progarm will be executed directly, without a shell.
On Linux the arguments are passed directly to execvp and no escaping
is necessary. On Windows we construct a command string using escaping
with the default Windows command-line argument parsing method described
at https://docs.microsoft.com/en-us/cpp/cpp/parsing-cpp-command-line-arguments.
Apart from avoiding the issue of argument escaping, passing an array
and bypassing shell has the advantage of allowing proper signal
delivery to the opened process (rather than the shell).
The actual leak is observed in ext/pdo_mysql/tests/bug_74376.phpt.
The persistent connection leaks because a refcount decrement on a
result is missed. The refcount decrement is missed because
free_result_contents is used, rather than free_result.
Looking at other uses of free_result_contents, it looks like they
could also suffer from this problem. Apart from one case,
free_result_contents is always used to release the result entirely
(I've adjusted the one differing case to only free meta), so I'm
moving most of the logic from free_result into free_result_contents.
The only difference is now that free_result will skip_result first.
If we're testing ini settings that cause startup failures, we'll
never get to that SKIPIF block...
Also change settings2param to return a value instead of modifying
(to a different type no less!) in place.
The strcoll function is defined in the C89 standard and should be
on today's systems always available via the <string.h> header.
https://port70.net/~nsz/c/c89/c89-draft.html#4.11.4.3
- Remove also SKIPIF strcoll check in test
The hash extension is always available since PHP-7.4. The symbol
HAVE_HASH_EXT is kept for BC reasons and removed in PHP-8.0.
This patch also removes the PHAR_HASH_OK since it is no longer
relevant.
In particular, make sure that everything using zmm is released
before zmm is shut down. phpdbg currently gets away with this,
because either a) its custom handlers are used and no auto-free
happens or b) the system allocator is used and no auto-free happens.
With the tracking allocator for asan this no longer works.
In this case we will use the system allocator, but still remember
all allocations and free them the same way that Zend MM does. This
allows us to accurately model leak behavior.
Enabled using USE_ZEND_ALLOC=0 USE_TRACKED_ALLOC=1.
As of libenchant 1.6.0 there is enchant_get_version()[1], so we use it,
if available, to show the actual enchant version in the PHP info. We
also drop the fake ENCHANT_VERSION_STRING altogether, but stick with
showing version 1.5.x if at least HAVE_ENCHANT_BROKER_SET_PARAM is
defined.
Future scope: we may consider requiring enchant 1.6.0 (or later), since
this has been released in April 2010, and likely is available
everywhere.
[1] <https://www.abisource.com/projects/enchant/>
Streams will be freed in an unpredictable order during shutdown.
Ignore explicit calls to php_stream_close() entirely to avoid
use-after-free -- instead let the stream resource destructor
deal with it. We have to account for a few special cases:
* Enclosed streams should be freed, as the resource destructor
will forward to the enclosing stream.
* Stream cookies also directly free streams, because we delegate
to the cookie destruction if one exists.
* Mysqlnd also directly frees streams, because it explicitly
removes stream resources (because mysqlnd!)
Some headers were checked multiple times in the main configure.ac file
and in the bundled extensions or SAPIs themselves. Also many of these
checks are then used accross other extensions or SAPIs so a central
configure.ac makes most sense for these checks.
This patch removes warnings when lexer and parser files are already
generated. For example, when downloading a PHP release. Realistically,
it is not so trivial to rebuild lexer and parser files without removing
them first. And considering that tarballs don't have cleaning tools
such as Git, this would require manual removali.
Related also to https://bugs.gentoo.org/593278
When tests are run concurrently using the -j option there might be a
change of failure and writing to the same directory from both tests.
Use test filenames for generated temp dirs
It seems that this code has a peculiar interpretation of "len",
where it actually points to the last character, not one past it.
So we need +1 here for that extra char and another +1 for the
terminating null byte.
This test is failing on AppVeyor almost all of the time, so splitting
it seems appropriate. This also allows us to rid php_cli_server_stop()
which was only used by this test case.
We add PHP bindings for libgd's features to read TGA files, which are
available as of libgd 2.1.0.
As PHP's bundled libgd doesn't yet include the respective features of the
external libgd, we add these.
Since TGA has no easily recognizable file signature, we don't add TGA
support for imagecreatefromstring() or getimagesize() and friends.
For some reason, when using GCC with address sanitizer, dlopen
is available without -ldl, but dlsym still needs it. Explicitly check
dlsym so we add the library.
symtable_cache_ptr now points to the first unused symtable_cache
entry, rahter than the last used one. This avoids taking a pointer
to the minus first element of the array, which is UB. Instead we
take a pointer to the end plus one, which is not UB.
The arena allocator has no problem with zero-size allocations (and
will just return the same pointer for each in this case), so just
do that to avoid null pointers.
The use of gettimeofday in time_sleep_until is technically incorrect, it's not possible
to use gettimeofday in this way reliably on any platform: It relies on operating system
global structures, which may be modified by any other process on the system at any time.
While in practice, users may be ignoring this flaw, it entirely depends on the other software
running on the system to which the application is deployed, there is no possible way to write
a test that will always pass on any system, therefore it must be marked XFAIL.
I've refactored based on the original bug report, related bugs, and commits
to php-src that were related to them.
It is supposed to be testing windows specific behaviour related to non-blocking pipes,
nevertheless the test runs everywhere.
As this test doesn't check differences in times before and after
operations, it doesn't actually verify anything and just wastes
30 seconds of test time.
As we already have other tests that do check how stat times change,
I'm removing it entirely.
This reverts commit bdcef51bcb.
It seems that pkg-config support for libargon2 is still flaky:
* No pc file on Alpine.
* Custom builds of released libargon2 versions create a broken
pc file. This is fixed in master, but not released.
Go back to the old detection code for now.
libcurl may link against 7 different ssl libraries, all of which
are thread-safe -- apart from openssl, of course. We check for
openssl and register locking callbacks in that case, but we should
not warn if any other library is used.
These tests actually check that no file with a name containing a NUL is
created by the GD image output functions. This is superfluous, since
it is sufficient to check that the function failed, and that an
appropriate warning has been raised.
We also add missing nullbyte injection tests.
On Mac thread_local and __thread are not ABI compatible, in addition, thread_local
comes with additional overhead, __thread seems to be the most suitable linkage to use
regardless of c++/c
Keep track of delayed variance obligations and check them after
linking a class is otherwise finished. Obligations may either be
unresolved method compatibility (because the necessecary classes
aren't available yet) or open parent/interface dependencies. The
latter occur because we allow the use of not fully linked classes
as parents/interfaces now.
An important aspect of the implementation is we do not require
classes involved in variance checks to be fully linked in order for
the class to be fully linked. Because the involved types do have to
exist in the class table (as partially linked classes) and we do
check these for correct variance, we have the guarantee that either
those classes will successfully link lateron or generate an error,
but there is no way to actually use them until that point and as
such no possibility of violating the variance contract. This is
important because it ensures that a class declaration always either
errors or will produce an immediately usable class afterwards --
there are no cases where the finalization of the class declaration
has to be delayed until a later time, as earlier variants of this
patch did.
Because variance checks deal with classes in various stages of
linking, we need to use a special instanceof implementation that
supports this, and also introduce finer-grained flags that tell us
which parts have been linked already and which haven't.
Class autoloading for variance checks is delayed into a separate
stage after the class is otherwise linked and before delayed
variance obligations are processed. This separation is needed to
handle cases like A extends B extends C, where B is the autoload
root, but C is required to check variance. This could end up
loading C while the class structure of B is in an inconsistent
state.
ext/gd/gd.c seems to return null for wrong parameter count, false for
error, and long otherwise. See https://php.net/imagecolorallocate
In php 8.0, the MAY_BE_NULL can probably be removed.
This should only have one r.
http://php.net/get_parent_class mentions that it returns string|false
Also, I'm not 100% sure if this is right - should this include
MAY_BE_NULL to handle the case of too many parameters?
(then again, get_called_class() also returns null in the global scope)
```
php > var_export(get_parent_class(2,3));
Warning: get_parent_class() expects at most 1 parameter, 2 given in php
shell code on line 1
NULL
```
These regularly fail due to port clashes, though we don't seem to
use those ports in any other tests.
As we don't care about the specific port that is used here, pass
zero so that a random ephemeral port is used.
Instead of checking GetBinaryType() for each file, we do a much cheaper
pre-check whether the filename extension matches .exe or .com, and call
GetBinaryType() only in this case. For BC we also report .bat and .cmd
files as executables again.
The patch has been provided by @weltling.
For two reasons: First, it's generally cheaper to match a regex than
perform a stat (especially on Windows). Second, it will not fail on
concurrent modification of a directory in parts that are not matched
by the pattern, such as the spurious failure in ext/zip/tests/bug72660.phpt.
We weren't able to do this in 7.1 because the deprecation notice
may be converted to an exception and __toString() can't throw,
which means that it ultimately become a fatal error. This issue
is resolved now, so we can mark the method as deprecated.
RFC: https://wiki.php.net/rfc/tostring_exceptions
And convert some object to string conversion related recoverable
fatal errors into Error exceptions.
Improve exception safety of internal code performing string
conversions.
These regularly fail due to port clashes, though we don't seem to
use those ports in any other tests.
As we don't care about the specific port that is used here, pass
zero so that a random ephemeral port is used.
It should be possible to skip any of these (and use the ini configured
defaults) by passing null, independently of strict_types settings.
Noticed while working on GH-4227.
The test directory is often used for temporary files, let's use
the source directory, which should be more stable. If that doesn't
work we need to create a temporary directory.
This is our most common intermittent failure. I don't understand
why it happens and can't reproduce it, but defining an explicit
sort order seems worth a shot...
Don't generate explicit warnings for these in two places, use usual
error handling mechanism. Additionally suppress a number of warnings
if the server has gone away.
php_win32_signal_system_ctrl_handler() is called from a kernel thread,
so the former initialization of `vm_interrupt_flag` has no effect,
since it is defined as thread-local. This is, however, not necessary,
since the CTRL signal handling is supposed to work only for the main
thread anyway. We therefore change `vm_interrupt_flag` and the related
variables to true globals.
This also allows us to unmark the respective test case as XFAIL.
Furthermore, `vm_interrupt_flag` is declared as `zend_bool *`, so we
better treat it such.
Causes build failure on release+zts azure build. I'm rewriting this
code to separate the if/else handling, because they don't really
have anything in common anyway...
Fails for me locally due to different number of warnings with
different messages. Rather than adding more wildcards I'm dropping
this test entirely, as it doesn't seem to test anything particularly
useful.
Scalar FFI values now should be accessed through special "cdata" property.
$x = FFI::new("int");
$x = 42;
should be changed into
$x = FFI::new("int");
$x->cdata = 42;
This allows writing
array_merge(...$arrays)
instead of
array_merge([], ...$arrays)
and is in line with similar changes to array_push() and array_unshift()
in PHP 7.3.
Closes GH-4175.
Even though we don't need it at runtime, add the BIND_IMPLICIT
flag to BIND_STATIC as well, so we can distinguish this case in
type inference.
This fixes a JIT miscompile in arrow_functions/002.phpt.
The num_args does not include variadics, so a "zero-arg" function
may accept additional arguments through that. No functions seem
to be affected right now, but they will be after #4175.
This is a minimal variance implementation: It does not support any
cyclic type dependencies. Additionally the preloading requirements
are much more restrictive than necessary. Hopefully we can relax
these in the future.
Fixes bug #76451, and more importantly lays necessary groundwork for
covariant/contravariant types. Bug #76451 is just an edge case, but
once covariance is introduced this will become a common problem instead.
Add XLEAK section, this is made necessary by the support
for valgrind tools other than memcheck.
XLEAK is to a leaking test what XFAIL is to a failing test.
Tidy up summary a little to exclude Expected Fail and Expected Leak
when there are no expected fail/leaks
Opcache stores `opline->handler`s in shared memory. These pointers are
invalid, if the main PHP DLL is loaded at another base address due to
ASLR. We therefore store the address of `execute_ex` in the mmap base
file, and check on startup whether it matches its current address. If
not, we fall back on the file cache if enabled, and bail out otherwise.
This still does not address cases where the opline handler is located
inside of another DLL (e.g. for some profilers, debuggers), but there
seems to be no general solution for now.
This enhances the makidst script:
- integrate both snapshot and makedist scripts together
- add help and options
- generated files are created in the php-src repository directly
- other minor enhancemenets such as CS fixes
- functionality moved from the Makefile to only shell script
- Add missed patching of the Zend Parsers to the main build step
- Add all *.tmp files to gitignore
This syncs the configuration option of the litespeed module to be
compliant with the approach of:
--with-foo (when something additional needs to be installed such as
libfoo)
--enable-foo (when extension/sapi doesn't have dependencies on some
library)
There really isn't any practical difference between these two, but repo
currently uses *.tiff files more than 8.3 variant filenames of *.tif.
Also these aren't part of the test.
Changes:
- executable from any location (for example, project root)
- some minor common shell scripts CS fixes
- error reporting done based on the presence of the parser file
This removes the example script in favour of the manual:
https://www.php.net/manual/en/book.pgsql.php
Patch for integrating these functions as an example has been submitted
also to PHP manual via edit.php.net.
`curl_version()`[1] (of ext/curl) makes `curl_version_info()`[2] (of
libcurl) available to PHP userland. The latter requires to pass an
`age` argument which usually is `CURLVERSION_NOW`, so that the
information returned by the runtime matches the declarations used
during compile time. For C programs it is simply necessary to pass
this information, and in rare occasions it might make sense to pass
something else than `CURLVERSION_NOW`. curl.h notes:
| The 'CURLVERSION_NOW' is the symbolic name meant to be used by
| basically all programs ever that want to get version information.
For the PHP binding, using a newer `age` than available at compile time
will neither provide the PHP program more information, nor would using
an older `age` have tangible benefits.
We therefore deprecate the useless `$version` parameter, and if it is
passed nonetheless, we use `CURLVERSION_NOW` instead of the supplied
value, and raise a warning.
[1] <https://www.php.net/manual/en/function.curl-version.php>
[2] <https://curl.haxx.se/libcurl/c/curl_version_info.html>
Changes to simplify managing libgd docs a bit:
- Move copyright statement and licensing info to common redistributable
bins file. All previous notices left intact and updated with the
upstream info.
- Remove unsynced and difficult to track README file compared to
upstream README.md file available at https://github.com/libgd/libgd
Normalization include:
- Use dnl for everything that can be ommitted when configure is built in
favor of the shell comment character # which is visible in the output.
- Line length normalized to 80 columns
- Dots for most of the one line sentences
- Macro definitions include similar pattern header comments now
Since all current ODBC test cases use the same DSN, there may be
conflicts when running tests in parallel. We prevent this by adding a
CONFLICTS marker which might be unnecessarily general, but get's the
job done, and still can be refined later.
- All credits, license info preserved, only merged together to not have
to many places to edit.
- Changelogs integrated in the readme since they are not used anymore.
This filter is available as of libgd 2.1.0 which is our bare minimum for
external compilation of ext/gd.
The scatter filter works by iterating over all pixels in the image and
shifting them randomly based on two modifier (`plus` and `sub`) values:
dest_x = (int)(x + ((rand() % (plus - sub)) + sub));
dest_y = (int)(y + ((rand() % (plus - sub)) + sub));
Additionally the scatter filter also supports by only shifting pixels where
the current pixel being iterated is one or more colors, allowing the scatter
filter to only effect solid colors in part of an image.
Note, due to the nature of randomness and implementation, pixels who were
shifted ahead of iteration will be shifted once more and therefore the
bottom right of an image may contain a slight scatter effect due to this.
Make sure to always fetch the RHS of a list assignment first, instead
of special casing known self-assignments, which will not detect cases
using references correctly.
As a side-effect, it is no longer possible to do something like
byRef(list($x) = $y). This worked by accident previously, but only
if $y was a CV and the self-assignment case did not trigger.
However it shouldn't work for the same reason that byRef($x = $y)
doesn't. Conversely byRef(list(&$x) = $y) and byRef($x =& $y)
continue to be legal.
The vcsclean script is really only a wrapper for a git clean command.
Developers should use the more proper and clear native Git command
directly instead:
`git clean -Xfd`
Make the check covariant (insofar as it is allowed now, i.e.
nullability and iterable) and call it with appropriate argument
order for both parameter and return types.
This makes it simpler to extend to full variance support.
`CURLPIPE_HTTP1` is deprecated and has no effect as of cURL 7.62.0[1].
We therefore deprecate the PHP constant as well, and trigger a warning
that it is no longer supported, if used against cURL 7.62.0 and up.
[1] <https://curl.haxx.se/libcurl/c/CURLMOPT_PIPELINING.html>
On A72, google-benchmark measure before and after the patch:
--------------------------------------------------------------
Benchmark Time CPU Iterations
--------------------------------------------------------------
BM_hash_init_before 43.6 ns 43.6 ns 16052937
BM_hash_init_after 27.0 ns 27.0 ns 25877296
Patch written by Ali Saidi <alisaidi@amazon.com>
and Sebastian Pop <spop@amazon.com>
The execution time goes from 4.388s down to 0.563s on a Graviton A1
instance for the benchmark:
function reverse_strings() {
$a = "foo";
for ($i = 0; $i < 100000; $i++) {
strrev($a);
$a .= "o";
}
}
This patch simplifies line endings tracked in the Git repository and
syncs them to all include the LF style instead of the CRLF files.
Newline characters:
- LF (\n) (*nix and Mac)
- CRLF (\r\n) (Windows)
- CR (\r) (old Mac, obsolete)
To see which line endings are in the index and in the working copy the
following command can be used:
`git ls-files --eol`
Git additionally provides `.gitattributes` file to specify if some files
need to have specific line endings on all platforms (either CRLF or LF).
Changed files shouldn't cause issues on modern Windows platforms because
also Git can do output conversion is core.autocrlf=true is set on
Windows and use CRLF newlines in all files in the working tree.
Unless CRLF files are tracked specifically, Git by default tracks all
files in the index using LF newlines.
- More common filename accross the PHP repository
- Additionally, this patch replaces some legacy form feed (FF or ^L)
characters (for printers) to LF (\n) newline character.
This script hasn't been used since using PEAR as a package manager for
PHP extensions since it is using Makefile.in as an indicator if the
current directory is extension. Instead of this script extensions can
be packaged differently and more properly with either current PEAR
or with some other manual method.
These have been inadvertently dropped when changing the test suite to
not require ext/interbase anymore, so we add them back.
We also change the required environment variable names to match the
usual PDO names. Particularly, we replace `PDO_FIREBIRD_TEST_HOSTNAME`
and `_DATABASE` with the more flexible `PDO_FIREBIRD_TEST_DSN`.
- This renames the environment variables to: PDO_FIREBIRD_TEST_XXX to be in line with other PDO test suites
- Adds an skipif.inc file that skips the tests if no database is set
- The test suite can now be run without PDO_FIREBIRD_TEST_HOSTNAME
$ set PDO_FIREBIRD_TEST_HOSTNAME=localhost
$ set PDO_FIREBIRD_TEST_DATABASE=C:\dev\php.fdb
$ nmake test TESTS=ext/pdo_firebird/tests
I have no idea how I managed to get this to work, but there is 4 new environment variables that
can be specified for the connection:
- PDO_FIREBIRD_USERNAME
- PDO_FIREBIRD_PASSWORD
- PDO_FIREBIRD_HOSTNAME
- PDO_FIREBIRD_DATABASE
(See testdb.inc for more info)
The only way I could get this to work on my local machine was with a DSN like:
- firebird:dbname=localhost:C:\php.fdb
Hence why the 'hostname' can be specified.
There is also a bit mess in regards to cleanup of this extension tests, but I really do not want to dwell down there, if someone dares to do so, then please be my guest.
TSRM configuration header file was once created by separate autoconf
build system for TSRM and is with the current code not directly needed
like this anymore.
The MS docs on `SQLColAttribute()`[1] state regarding the
`NumericAttributePtr` parameter:
| Please note that some drivers may only write the lower 32-bit or
| 16-bit of a buffer and leave the higher-order bit unchanged.
| Therefore, applications should initialize the value to 0 before
| calling this function.
[1] <https://docs.microsoft.com/en-us/sql/odbc/reference/syntax/sqlcolattribute-function?view=sql-server-2017>
Due to former restrictions of the libcurl API, curl multipart/formdata
file uploads supported only proper files. However, as of curl 7.56.0
the new `curl_mime_*()` API is available (and already supported by
PHP[1]), which allows us to support arbitrary *seekable* streams, which
is generally desirable, and particularly resolves issues with the
transparent Unicode and long part support on Windows (see bug #77711).
Note that older curl versions are still supported, but CURLFile is
still restricted to proper files in this case.
[1] <http://git.php.net/?p=php-src.git;a=commit;h=a83b68ba56714bfa06737a61af795460caa4a105>
The buildmk.stamp file has been created by the *nix build checking step
to run the check step only once. Instead of poluting the project root
directory, the stamp file can be also omitted. Performance difference is
very minimal to not justify having the stamp check at all today anymore.
This patch integrates the buildcheck.sh to buildconf script directly.
This was once part of TSRM but then got refactored into the windows
implementation win32/readdir.h directly. Instead of including such files
directly code should use zend_virtual_cwd.h which is already part of the
php.h file.
The force option used to only remove the Autoconf cache files. Now it
also removes entire configure file so the configure file always gets
rebuilt when needs be done.
Additionally, the buildconf now also accepts the shorter version of the
force option:
buildconf -f
Commit 4e7064d173 removed the usage of
`aclocal.m4`. When using Git repositories, many times cleaning of the
generated files is not done prior to running phpize or buildconf. For
example:
git clone git://github.com/php/php-src
cd php-src
git checkout PHP-7.3
./buildconf
./configure
git checkout PHP-7.4
./buildconf # -> warnings
./configure # -> errors
To not accidentally include `aclocal.m4` file in the generated configure
this enhances build system experience a bit more by removing aclocal.m4
file prior to start building configure file using phpize or buildconf.
Add --enable-werror configure option and use it on Travis. It's not
possible to directly use CFLAGS, because it also affects configure
checks which often throw warnings.
We can't enable something similar for Windows builds at this time,
because they throw a lot more warnings.
We must not (try to) modify shared values, but rather have to use our
own copies, if unixified filenames are required on Windows. To avoid
excessive string duplication, we add checks whether the filenames are
already unixified (i.e. do not contain backslashes). To improve the
performance if we need to copy strings, we use do_alloca() and friends.
Besides generally being somewhat messy, the handling of unixified
filenames is still suboptimal performance-wise, but we leave this for a
future cleanup, and focus on fixing the issue at hand for now.
We also enable opcache.protect_memory for the AppVeyor CI.
The acinclude.m4 file is in a usual Autotools build processed with
Automake's aclocal tool. Since PHP currently doesn't use Automake and
aclocal this file can be moved into the build directory. PHP build
system currently generates a combined aclocal.m4 file that Autoconf
can processes automatically.
However, a newer practice is writing all local macros in separate
dedicated files prefixed with package name, in PHP's case PHP_MACRO_NAME
and putting them in a common `m4` directory. PHP uses currently `build`
directory for this purpose.
Name `php.m4` probably most resembles such file for PHP's case.
PHP manually created the aclocal.m4 file from acinclude.m4 and
build/libtool.m4. Which is also not a particularly good practice [1], so
this patch also removes the generated alocal.m4 usage and uses
m4_include() calls manually in the configure.ac and phpize.m4 files
manually.
- sort order is not important but can be alphabetical
- list of *.m4 files prerequisites for configure script generation
updated
- Moving m4_include() before AC_INIT also removes all comments starting
with hash character (`#`) in the included files.
[1] https://autotools.io/autoconf/macros.html
The xmlrpc library is a forked, modified and bundled into the PHP source
code. Instead of having a separate *nix build system files in the xmlrpc
library this patch removes unneeded build checks.
- AC_PROG_CC is not needed to be called by the extension itself since it
is already part of the main configure.ac and the phpize's
configure.ac.
- UNDEF_THREADS_HACK is not used in the current codebase.
- Symbols defined by XMLRPC_HEADER_CHECKS are not used in the current
code.
- Symbols defined by XMLRPC_FUNCTION_CHECKS are not used in the current
code.
- AC_PROG_RANLIB is already done by the TSRM's files and isn't needed
for the xmlrpc extension to work.
- AC_PROG_LN_S is already done by the PHP main configure.ac and also
isn't needed by the xmlrpc extension to be built.
Arbitrary DateInterval objects don't have well-defined comparison
semantics. Throw a warning and treat the objects as uncomparable.
Support for comparing DateInterval objects returned by
DateTime::diff() may be added in the future.
There is now only a single M4 macro in the legacy acinclude.m4 file. A
separate acinclude file was once used with a standalone Zend engine
building but with current build system this can be simplified a bit.
Checked and quickfixed with Markdown linter
- 80 columns line width (~)
- code highlighting
- ...
Some most obvious outdated content updated a bit more.
- [date_timestamp_get] adding coverage for date_timestamp_get alias to
DateTime::getTimestamp feature at PHPRio TestFest
- [date_timestamp_get] adding coverage for date_timestamp_get basic
feature at PHPRio TestFest
- Combine tests into single file
With recent transition to pkg-config/pkgconf this macro has been now a
simple wrapper for PKG_CHECK_MODULES and PHP_EVAL_INCLINE so it's better
to omit it altogether and use these two in the *.m4 code directly.
This reverts commit 7435ab3375.
It looks like we can't access EG(current_module) here in ZTS builds.
As the problematic asserts have now been disabled outside RC_DEBUG,
I'm just reverting this commit and leaving things as they were.
FOREACH_END_DEL() does not call the destructor by itself, we need to
explicitly call it. Due to the missing dtor call the functions were
not unregistered early, which resulted in later shutdown segfaults.
Changes:
- Joins build/build.mk and build/build2.mk files together since there
isn't any practical reason for having two different files with the
current build system.
- Makefile is now more portable. All special syntaxes are omitted, for
example, a conditional assignment operators `?=`. This makes buildconf
more useful on Solaris make derivative, so there is no longer need to
override make with gmake: `MAKE=gmake ./buildconf`.
- Suppressing autoconf and autoheader warnings is not needed anymore
with current build system. Instead, the option `-Wall` has been used
when running `./buildconf --debug` to get more useful debug info
about current M4.
By introducing a hook that is called whenever one of
internal_encoding / input_encoding / output_encoding changes, so
that mbstring can adjust it's internal state.
This also makes internal_encoding work with zend multibyte.
The `generated_lists` file is generated as a helper for build related
Makefile to include a list of *.m4 files prerequisites. When some of
these *.m4 files change, the configure script is regenerated when
buildconf is run. This can be simplified using dynamic environment
variable passed to the Makefile directly so it avoids another file from
being generated in the project root directory and shipping it with the
PHP release or creating a dedicated gitignore rule.
This is portable across all POSIX compatible makes So this patch
includes GNU Make, and everybody elses' make derivative support.
The HAVE_PCRE_JIT_SUPPORT check uses AC_RUN_IFELSE, which is not
available when cross-compiling. As a fallback, JIT support is enabled
based on CPU architecture. However, this may be wrong,
e.g. when the JIT the feature was not enabled in the pcre2 build.
Add a cache variable for the PCRE JIT feature to make it possible to
override the check.
* 'PHP-7.4' of git.php.net:php-src:
Remove not needed extension generated files gitignores
Add test for curl_version()
Fixed bug #77895
Zero sockaddr struct
Fix saproxy_property_write signature
Correctly destroy reference in ArrayObject sort
It seems like GCC 5 likes to warn about exactly those trivial strict
aliasing violations that it's not going to miscompile anyway and the
warnings are gone in newer versions. Just disable the noise.
Autoconf defines PACKAGE_* symbols:
- PACKAGE_NAME
- PACKAGE_VERSION
- PACKAGE_TARNAME
- PACKAGE_STRING
- PACKAGE_BUGREPORT
- PACKAGE_URL
and appends them to the generated config.h.in files. With AC_INIT change
via afd52f9d99 where package version, URL,
bug report location and similar meta data are defined, these
preprocessor macros are then non empty strings in the generated
configuration header file. When using phpize, PHP shares the config
files in extensions, warnings of redefined macros appear, such as:
- `warning: 'PACKAGE_NAME' macro redefined`
This patch now disables these non utilized symbols in the generated
config header files.
Better practice would be to include only API specific headers where
needed but this would require even more refactorings. Some extensions
such as pcre, pgsql, and pdo_pgsql solve this issue by undefining some
of these symbols before including the library configuration headers in
the code also. Because these symbols can be defined by any library which
uses Autotools.
Additionally, the unused PACKAGE_* symbols were cleaned for the bundled
libmbfl library and with this patch not needed undef code removed.
This is unnecessary convoluted... we just want to find the position
of the dot. Using php_strtok_r requires a copy of the string, and the
implementation also relied on the specific semantics of the scratch
space that strtok_r uses.
This one looks semi-legit, in case php_stream_eof() returns false
but php_stream_get_line() fails. Not totally sure this cannot
happen, so rewriting to check both conditions at once.
And fix incorrect variable shadowing in add_config_entry(). However,
the test doesn't hit this case, as it requires a nested array. I'm
not sure if it's possible to produce nested arrays from ini?
This code was reusing the _bailout variable from
SOAP_CLIENT_BEGIN/END_CODE(). As _bailout is not volatile, modifying
it after the setjmp call and then reading it back on return is
illegal. Use a separate local bailout variable instead.
This fixes the miscompile introduced by marking zend_bailout() as
noreturn.
As we are breaking out of the for loop setting i = ssa_vars_count
doesn't do anything, but it confuses GCC into throwing a no-overflow
assumption warning.
If there are no warnings, we may end up compiling more than 10
minutes without output, in which case the build times out. Use
travis_wait to avoid this.
readdir_r() is deprecated in modern glibc versions. readdir() is
thread safe in practice, as long as there are no concurrent accesses
on the *same* directory stream.
This is only relevant for the gcrypt backend, which is no longer in
use. I couldn't find any hard info on when exactly gcrypt support was
dropped, but nette is the default since 2.11.1. Most code that still
has the gcrypt handling is checking against < 2.12. In either case,
those versions were released in 2010/2011.
We're not using this anywhere, and it causes warnings, so I'm
dropping it. If we do end up needing this, it should be reintroduced
with a ZEND_ prefix.
We're not checking the return value and the NTS version of this
generates warnings. If we want to handle lock failures, we should
do a hard abort inside tsrm_env_lock() itself.
The substitution is already done in the CLI's config.m4 file. Current
SAPIs only provide one PHP_EXECUTABLE variable, i.e. PHP CLI so the one
in the configure.ac can be removed.
Checking for the strerror function presence is no longer needed since it
is part of the C89 standard [1] and can be safely assumed that all
current systems have it.
Check in the configure.ac and Windows defined symbol are still left
until the file library (libmagic) will be updated.
[1]: https://port70.net/~nsz/c/c89/c89-draft.html
php_filter_int (called via the constant FILTER_VALIDATE_INT) has the options min_range
and max_range. they allow the user to not only test if a value is a double but also if
the value is inside a specific range. php_filter_float (called via the constant
FILTER_VALIDATE_FLOAT) didn't provide this yet, making validation of numeric but
not-only-int values more complicated for the user.
this commits implements the options min_range and max_range for the function
php_filter_float to fix this inconsistency.
The `<loccale.h>` header file, setlocale, and localeconv are part of the
standard C89 [1] and on current systems can be used unconditionally.
Since PHP 7.4 requires at least C89 or greater, the `HAVE_LOCALE_H`,
`HAVE_SETLOCALE`, and `HAVE_LOCALECONV` symbols defined by Autoconf in
configure.ac [2] can be ommitted and simplifed.
The bundled libmagic (file) has also been patched already in version
5.35 and up in upstream location so when it will be patched also in
php-src the check for locale.h header is still left in the configure.ac
and in windows headers definition file.
[1] https://port70.net/~nsz/c/c89/c89-draft.html#4.4
[2] https://git.savannah.gnu.org/cgit/autoconf.git/tree/lib/autoconf/headers.m4
Omit the bundled libmagic files
The `<signal.h>` header file is part of the standard C89 headers [1] and
on current systems can be included unconditionally.
Since file requires at least C89 or greater, the `HAVE_SIGNAL_H` symbol
defined by Autoconf in Zend.m4 [2] can be ommitted and simplifed.
The bundled libmagic (file) also ommits the usage of HAVE_SIGNAL_H since
5.35 however current version in PHP is very modified 5.34 version and
will be refactored separately. Check for HAVE_SIGNAL_H is therefore
still done in the configure.ac.
Refs:
[1] https://port70.net/~nsz/c/c89/c89-draft.html#4.1.2
[2] https://git.savannah.gnu.org/cgit/autoconf.git/tree/lib/autoconf/headers.m4
The `<limits.h>` header file is part of the standard C89 headers [1]
and on current systems can be included unconditionally.
Since PHP requires at least C89 or greater, the `HAVE_LIMITS_H` symbol
defined by Autoconf in configure.ac [2] can be ommitted and simplifed
however due to bundled file library (libmagic) and timelib still using
it, the removal there was omitted and done only in Zend.m4 file.
Current bundled libraries libtime, oniguruma, and libmagic still include
partial `HAVE_LIMITS_H` usage and will be more refactored when this is
possible.
Refs:
[1] https://port70.net/~nsz/c/c89/c89-draft.html#4.1.2
[2] https://git.savannah.gnu.org/cgit/autoconf.git/tree/lib/autoconf/headers.m4
Since Autoconf 2.53 the AC_INIT call with only a single argument has
been made obsolete and now includes several other optional arguments to
make installation experience a bit better by providing program version
and links to the project in the `./configure -h` output. This patch also
updates win build version. The phpize.m4 AC_INIT has been updated with
the call without arguments.
For the latest versions, refer also to the Visual Studio version, not to
the exact toolset version. The latest Visual Studio versions are moving
fast and the method existed previously is not sutable to keep up
anymore. Instead of refering to the exact toolset, it refers to the
Visual Studio version and implies the latest toolset available there.
That is still not ideal, another situation can arise where VS version is
not changed, but the toolset version is. It might be better to refer to
it a generic way in the future or even omit at all, and improve the
internal compatibility checks instead.
This hardens the dynamic module loading by checking the linker compatibility
between the core and the dynamic module. This likely should be extended
for the CRT as well, as 2015, 2017 and 2019 versions of Visual Studio
all have same DLL name for the CRT.
The new dedicated docs directory has been introduced after a discussion
on GitHub[1].
Main issue it is addressing is the reduction of too many README files in
the project root directory. The new directory is dedicated for notes and
quick documentation files that either can't be put in the manual or wiki
pages or that relate to the php-src repository specifically and need to
live together with the source code.
Also the `docs` directory is by GitHub used for some repository
configuration files such as pull request templates, and contributing
documentation helper files that are integrated in the interface.
[1]: https://github.com/php/php-src/pull/3988
This patch joins two very much related pieces of docs together in a
single file dedicated to all sorts of contributing info.
Some more changes:
- Branches info copied from the current master branch
- LXR and bonsai info removed
- Duplicated info reduced a bit
- Security branch updated to 7.1
- Refactor intro for Git commit rules
- Updated README.GIT-RULES file usage in win32/build/confutils.js
- Refactored configure.ac
This patch refactors these macros to also checks for the required given
versions of bison and re2c.
- PHP_PROG_RE2C and PHP_PROG_BISON take optional args - minmimum version
required, and bison also excluded versions.
- Instead of caching values this uses manual checking and messaging
outputs.
- It looks like the minimum version of RE2C 0.13.4 is working ok so far.
The genfiles script improvements:
- Add make override in genfiles
- Move checkings from makedist to genfiles
- Refactored output messages
- Various minor enhancements
RPM specification file was introduced via
7c2f1384d4 for PHP to include official
RPM packages long time agon. With removal of the makerpm script via
3d51d4c90c and Linux repositories to
manage such updated and customized info in their repositories this file
is most likely not needed anymore.
* PHP-7.3:
Validate subject encoding in mb_split and mb_ereg_match
Validate pattern against mbregex encoding
SQLite3: add DEFENSIVE config for SQLite >= 3.26.0 as a mitigation strategy against potential security flaws
Abstract ctor signatures should always be respected by all children,
independently of whether it comes from an interface or an abstract
class. Previously abstract ctor signatures (if they didn't come from
an interface) were only checked to one level of inheritance.
There will only be one request on the CLI SAPI, so there is no
advantage to having a persistent PCRE cache. Using a non-persistent
cache allows us to use arbitrary strings as cache keys.
This joins two similar contextual information together in the main
CONTRIBUTING.md file.
Additionally:
- http links updated to https where possible
- sentences and sections refactored more so they make some sence. GitHub
pull requests are preferred way for contributing code.
- reorder paragraphs and contextual info
- s/mail list/mailing list
- PHP internals refactorings
- use canonical https://www.php.net
Motivation:
As an extension author, I want to speed up running tests in php <=7.3,
both locally and in CI (e.g. with valgrind).
This can be done by manually copying php 7.4's run-tests.php script
to replace the one generated by `phpize`
- list() doesn't work in php 7.0
- negative string offset doesn't work in php 7.2
If run-tests.php can be copied from php-src without any manual patches,
that would be the easiest.
Related to #2822 - I didn't see any discussion for/against compatibility
with older php versions
With this required Autoconf version is now defined only on two places:
- configure.ac
- scripts/phpize.m4
and additionally:
- Script can be run from other locations
- Synced CS and portability a bit
Locations for AST nodes are now tracked with the help of bison
location tracking. This is more accurate than what we currently do
and easier to extend with more information.
A zend_ast_loc structure is introduced, which is used for the location
stack. Currently it only holds the start lineno, but can be extended
to also hold end lineno and offset/column information in the future.
All AST constructors now accept a zend_ast_loc* as first argument, and
will use it to determine their lineno. Previously this used either the
CG(zend_lineno), or the smallest AST lineno of child nodes.
On the parser side, the location structure for a whole rule can be
obtained using the &@$ character salad.
Accept the two offsets directly, rather than doing length calculations
at all callsites. Also extract the logic to create a possibly interned
string.
Switch the split implementation to work on a char* subject internally,
because ZSTR_VAL(subject_str) is a mouthful...
Export zend_release_fcall_info_cache(). It is only necessary to
call it if the fcc may not have been used -- if it is passed to
zend_call_function() and friends, then they will take care of
freeing trampolines.
With Bison 3.0 some directives are deprecated:
- %name-prefix "x" should be %define api.prefix {x}
- %error-verbose should be %define parse.error verbose
Bison 3.3 also started emiting more warnings and since PHP souce parsers
are not POSIX compliant this patch fixes this as pointed out via
495a46aa1d.
This issue was mentioned in bug #73948. The PREG_PATTERN_ORDER
padding was performed without respecting the PREF_OFFSET_CAPTURE
flag, which resulted in unmatched subpatterns being either null or
[null, -1] depending on where they occur. Now they will always be
[null, -1], consistent with other usages.
If PREG_OFFSET_CAPTURE is used, unmatched subpatterns will be either
[null, -1] or ['', -1] depending on PREG_UNMATCHED_AS_NULL mode.
Instead of creating a new array like this every time, cache it inside
a global (per-request -- could make it immutable though).
Additionally check whether the subpattern is an empty string or
single character string and use an existing interned string in that
case. Empty / single-char subpatterns are common, so let's avoid
allocating strings for them.
When used with preg_match_all or preg_replace_callback(_array),
subpattern names can be used in the matches array many times.
Switch the subpat_names table to use zend_string, so we don't have
to allocate a new string every time. Also don't bother creating the
table if no $matches were passed.
This might be a regression for the case where preg_match() is used
with many trailing named subpatterns that are skipped in the result
array, but that seems rather contrived.
MYSQLI_ASYNC is also valid here, at least with mysqlnd. Rather than
using a random mode that is prone to failing once in a blue moon,
use a fixed invalid value.
We currently have a large performance problem when implementing lexers
working on UTF-8 strings in PHP. This kind of code tends to perform a
large number of matches at different offsets on a single string. This
is generally fast. However, if /u mode is used, the full string will
be UTF-8 validated on each match. This results in quadratic runtime.
This patch fixes the issue by adding a IS_STR_VALID_UTF8 flag, which
is set when we have determined that the string is valid UTF8 and
further validation is skipped.
A limitation of this approach is that we can't set the flag for interned
strings. I think this is not a problem for this use-case which will
generally work on dynamic data. If we want to use this flag for other
purposes as well (mbstring?) then it might be worthwhile to UTF-8 validate
strings during interning. But right now this doesn't seem useful.
- AC_CHECK_SIZEOF is now called mostly only in PHP_CHECK_STDINT_TYPES()
macro except for some parts checking for the 32 or 64 bit architecture.
- SIZEOF_CHAR removed since it is always 1
- ZEND_BIN_ID is now of a more logical pattern `BIN_48888` on 64bit
architectures and `BIN_44444` on 32bit instead of literal string
`BIN_SIZEOF_CHAR48888` on 64bit and `BIN_SIZEOF_CHAR44444` on 32bit.
The unneeded SIZEOF_CHAR part has been removed.
- XMLRPC_TYPE_CHECKS removed
- The `long long int` is the same as `long long` and redundant checks
removed accordingly.
- Removed PHP_CHECK_64BIT macro. Checking if current platform is 64bit
or not can be also done simply by using a check of the long type on
place. This removes redundant m4 macro PHP_CHECK_64BIT.
The php manual already includes introduction to the JSON extension. The
re2c and bison version required to build parser and lexer files have
changed so to move this info on a central place this removes the README.
The more proper place for shell scripts dedicated for development, and
releasing PHP should be the scripts/dev directory. Having a cleaner root
project directory helps find the main README.md and files relevant to
install PHP.
These scripts are also used by the release managers mostly who create
release packages and aren't used often by the majority of developers
working on and installing PHP.
Changes:
- Added a short introduction what this script does
- Added usually the expected -h and --help options with explanation
for a reason for this script and its usage.
- Messages changed a bit so the PHP installation procedure becomes
simpler without needing to constantly remind the reader what to run
and what not in the documentations and installation instructions.
- cd into current working directory of the buildconf (this enables
running the script from other locations and inside other scripts).
- check if make exists
The `make install` step touches files outside of the source directory
and might cause confusions with people. It should be mentioned in more
details elsewhere in the installation chapter instead and how to
override it.
Some of these are entirely unnecessary. Some specify an associativity
for operators where the concept doesn't make sense and has no effect:
Explicit associativity is only meaningful for binary operators. For
unary operators and constrained operations like instanceof or assignment
the associativity specification has no effect and only serves to confuse.
For example assignments were specified as %left, even though they will
actually behave as right-associative due to the variable requirement on
the LHS. Similarly instanceof was specified as %nonassoc, but it will
behave as left-associative, due to the restrictions places on the RHS.
Changes:
- AC_TYPE_SIZE_T called on only one place (configure.ac)
- AC_FUNC_ALLOCA called on only one place (configure.ac)
- AC_TYPE_UID_T called on only one place (configure.ac)
- HAVE_STRSTR removed since strstr is part of C89 standard [1]
- Remove checks for strtol and strpbrk
- Checking for the presence of perror function is not needed anymore
since it is part of C89 standard and PHP calls it unconditionally.
- Checking for functions strdup, setenv, strerror, and memmove done only
on one place (configure.ac)
- outdated check for snprintf removed
1: https://port70.net/~nsz/c/c89/c89-draft.html
Function strftime is part of the C89 standard [1] and current systems
don't need to check for its presence anymore.
Additionally, checks for strftime function in tests have been removed
since the PHP strftime function is now always available.
1: https://port70.net/~nsz/c/c89/c89-draft.html
- Sync help output using heredoc
- Add extension building instructions
- Building with phpize is preferred option. Mention also tests
- Refactor print_success()
It doesn't support method of internal classes invoked through C object or class handlers.
This commit prevents crash described at bug #77708, but doesn't implement the desired behavior.
This variant of crc32 is heavily used by storage systems, such as iSCSI, SCTP,
Btrfs, ext4, and is increasingly being used in API (such as Google Cloud
Storage, and Apache Kafka).
This check introduced in 2004 is wrong and removed. A HEAD request with
curl does *not* cause this error code - only if you make a regular GET
request but tell curl to send a HEAD using CURLOPT_CUSTOMREQUEST and
then you've asked for it.
You do a proper HEAD request with curl by setting the CURLOPT_NOBODY
option to 1L.
This was the case in 2004. This is still the case in 2019. This is also
documented in libcurl documentation.
This check hides the possibly serious error when this error code is
genuinely and correctly returned by curl because the transfer was
truncated and ended up partial.
As can be seen, I objected to this change already in the original bug:
https://bugs.php.net/bug.php?id=27341
GetFinalPathNameByHandleW is given a file handle to a symbolic link,
rather than one to the target, and therefore returns an incorrect path.
Fix symlink with relative path and add test
- remove some outdated and not used macro calls
- remove some unused variables
- Remove not needed comment from tokenizer config.m4
- Remove not needed comment
- remove not needed local variables for editors and syntax highlighting
- Initial refactorings of the README.md file with some more user
oriented information about the php-src repository and PHP in general.
- SVG logo added
- README.WIN32-BUILD-SYSTEM file contents moved to README.md
- INSTALL file contents moved to README.md
- Windows releases now instead of install.txt file ship with the
README.md.
- A brief installation section added with help options mentioned for the
configure and make.
- Also mentioned the `-j` option for build and tests
Co-authored-by: Javier Eguiluz <javier.eguiluz@gmail.com>
Co-authored-by: Gabriel Caruso <carusogabriel34@gmail.com>
Co-authored-by: Matthias Kaschubowski <nihylum@gmail.com>
The AC_CHECK_TYPE was refactored in more recent versions of Autoconf
and the call with two arguments is obsolete and not recommended anymore.
This patch also refactors some leftovers of using ulong and uint which
are not standard nor common usages of types in C.
The ulong can be used as zend_ulong and uint usage is actually
`unsigned int`.
The usage of HAVE_ULONG removed since it is not used in current code
base.
Legacy edgecase for some legacy HPUX systems removed:
- sys/stream.h header is not checked and the HAVE_SYS_STREAM_H is
not defined with current build system.
- flags are unsigned int
- max_allowed_packet changed to unsigned int
Signed shift of 31 for int and 63 for long is flagged as undefined
behavior by UBSan (-fsanitize=undefined) and seems to be indeed so
according to the standard.
The patch converts such cases to use unsigned.
Changes:
- PHP_TM_GMTOFF removed
- HAVE_TM_GMTOFF replaced with HAVE_STRUCT_TM_TM_GMTOFF
- HAVE_TM_ZONE replaced with HAVE_STRUCT_TM_TM_ZONE
- HAVE_TZNAME removed
The PHP_TM_GMTOFF macro can be replaced with Autoconf's AC_CHECK_MEMBERS
that defines the HAVE_STRUCT_TM_TM_GMTOFF symbol instead of the
HAVE_TM_ZONE.
The HAVE_TZNAME symbol is not used in current code. The obsolete
HAVE_TM_ZONE symbol has been replaced with more proper
HAVE_STRUCT_TM_TM_ZONE. These are defined by the AC_STRUCT_TIMEZONE
macro.
The AC_PROG_CC_C_O macro checks if compiler can use both -c and -o
options together and if not it defines the NO_MINUS_C_MINUS_O symbol.
It is not used in current codebase and therefore removed.
Checking for int32_t and uint32_t types is already done by the
PHP_CHECK_STDINT_TYPES m4 macro.
The PHP_CHECK_TYPES m4 macro has been once used by the mysqlnd and has
been removed via 14caf174ff and
additionally, the internal unused macro _PHP_DEF_HAVE_FILE has been
removed.
Additionally, the unused PHP_HAVE_STDINT_TYPES symbol has been removed.
Autoconf 2.59d (released in 2006) [1] started promoting several macros
as not relevant for newer systems anymore, including the `AC_HEADER_STDC`.
This macro checks if given system has C89 compliant header files such
as `<string.h>`, `<stdlib.h>`, `<stdarg.h>`, `<float.h>`,... and defines
the `STDC_HEADERS` symbol [2]. Case is that current systems should be
well supported with at least C89 standard headers [3].
Given headers are still additionally checked with the `AC_PROG_CC`
macro, yet not needed anyway.
Additionally, the HAVE_MEMCPY check has been removed. The memcpy
function is standardized by C89 and later.
Refs:
[1] http://git.savannah.gnu.org/cgit/autoconf.git/tree/NEWS
[2] https://www.gnu.org/software/autoconf/manual/autoconf-2.69/autoconf.html
[3] https://port70.net/~nsz/c/c89/c89-draft.html#4.1.2
Git can track executable (0755) and non-executable (0644) file modes.
This patch fixes file permissions in the php-src repository according to
the predefined executable files with 0755 permissions (shell scripts)
and all others with 0644 permissions.
`_php_image_output()` is only ever called for `PHP_GDIMG_CONVERT_WBM`,
`PHP_GDIMG_TYPE_GD` and `PHP_GDIMG_TYPE_GD2`. All other image types
solely use the more flexible `_php_image_output_ctx()`.
This code is not compatible with PHP 7.0. The fact that nobody
complained that the mysqli embedded server functionality doesn't
build anymore seems like a strong signal that we can drop it...
The relax NG schema is fetched from a remote server. We could
include it in the repo, but it's 500KB. A minimized test case would
be better.
Also, apparently this bug has never been reported upstream and just
patched in the libxml we build on Windows?
The snprintf function is part of the C99 standard and newer systems in
most cases all support it as defined in the standard. However, some old
Windows and HP-UX systems the function behaves differently. These checks
were also removed and PHP now uses a replacement for the snprintf
function. With gradual transition to C99 usage as a minimum requirement,
it will also be able to be replaced to system's snprintf function
directly.
Additionally in this context the unused HAVE_VSNPRINTF and check for
vsnprintf have been removed. PHP uses its own vsnprintf implementation
for now until more reliable C99 compliant function can be used from the
C libraries.
The PHP_EXTENSION macro was used before the introduction of the updated
build system in the 9d9d39a0de. The
extensions at that time possibly still used the Makefile.in and Automake
and the PHP_EXTENSION macro has been replaced with the PHP_NEW_EXTENSION
macro.
Today, the once deprecated macro can be removed in favor of only
PHP_NEW_EXTENSION macro.
The sprintf function has been normalized to php_sprintf via
61364b5bb1.
This patch removes the checks to make a custom sprintf function
The ZEND_BROKEN_SPRINTF has been removed and the
hardcoded #define zend_sprintf sprintf is used.
The php_sprintf and zend_sprintf are now symbols to sprintf.
This patch now removes the custom PHP definitions of the php_sprintf and
zend_sprintf functions in favor of the C99 sprintf which is also
standardized in C89 already. Once, on some systems sprintf didn't behave
in same way.
calendar.c includes ISO-8859-8 encoded Hebrew characters, which may
cause compile errors, and is causing issues when saving file as UTF-8.
This patch replaces characters with appropriate escape sequences.
Following 99fdf5916e.
Opcode order changes in 7.4 and the EXT_STMT is now declare the
DECLARE_ANON. Fix this by returning the opline from compile_class_decl
to avoid any fragile opcode searching.
Context: The ext/standard/tests/file/popen_pclose_error-win32.phpt
test often fails under parallel testing, because the "is not recognized
as an internal or external command" message doesn't actually have a
guaranteed position in the output.
While looking into this, I noticed that this test on Windows tests
something very different (invalid comand) than on Linux (invalid mode).
Here I'm adjusting the Windows popen implementation so it immediately
fails on a `rw` mode, just like it does on Linux.
If someone has a better patch, please merge it ASAP, this appears to be correct as I and Nikita originally thought.
Revert "Revert "zend_get_call_op ignoring compiler flags zend_get_call_op will ignore ZEND_COMPILE_IGNORE_USER_FUNCTIONS and ZEND_COMPILE_IGNORE_USER_FUNCTIONS, breaking the intention of these flags""
This reverts commit 0bbbd0f9e7.
This was once used by the ext/mysql extension named as MYSQL_LIB_CHK.
The call got removed via fd1578c196. In
PDO this is not called anymore and can be removed.
These tests are pointless (they don't even have anything to do
with resources, they just use get_resource_type() as a peculiar way
to create a string), and cause unnecessary IO dependency.
By using a separate file for each test. Also drop some unnecessary
SKIPIF checks for file creation. We assume that files can be created in
the test directories all over the place, so don't check for it here.
If a test conflicts with "all", then no other tests may be run in
parallel. This is needed for windows_mb_path tests, which rely on
the console codepage, which is shared across all parallel workers.
Also add support for comments in the CONFLICTS section/file.
If tests are run in parallel, files may be created and deleted in
the directory. Create a separate directory just for this test to
avoid this dependence.
Inherit the main stdin/stdout/stderr streams into the worker processes.
We don't talk to workers via these pipes, so there shouldn't be a
problem doing this.
Do not include unbound anonymous classes in get_declared_classes().
Note that earlier PHP versions would include the anonymous class in
get_declared_classes(), and return false until the class was bound,
but would not crash.
Fixes a bug in test added in
5661feb1ef, by checking that the
relevant SO exists.
Also adds an additional test that is based on zend_printf, and
should work on both Linux and Windows.
Compiler failures are not side-effect free, they may leave behind
parts of the file that were successfully compiled before the
failure occurred. The preloading code is not prepared to deal with
this.
At the time of this commit, there is a dedicated folder for development
related tools and such scripts might fit better there to not bloat the
project root directory too much.
Move snapshot to scripts/dev/snapshot
As of PHP 7.3.0, case-insensitive constants are deprecated. We catch
up on this with regard to ext/com_dotnet, which allows to import
constants from typelibs, by triggering a deprecation notice whenever
`com_load_typelib()` is called with `$case_sensitive` being `false`,
and whenever `com.autoregister_casesensitive` is set to `false`,
regardless of whether there are actually constants in the typelib which
would be imported.
During preloading, try to resolve all property types to CEs. Add a
flag that tracks this. If not all property types can be resolved,
then the class is not eligible for preloading.
This avoids writing this cache at runtime, which is illegal if
preloading is used.
Not every serialize/unserialize function actually belongs to the
Serializable interface, but I think it's not a problem to assign
these anyway -- whether they are used ultimately depends on whether
Serializable is implemented.
Alternatively it might make sense to just drop these entirely. I
don't think this is performance critical functionality.
Increased the coverage lines 587-595 in file ext/posix/posix.c
User Groups: PHPVale PHPSP PHPDF PHPWomenBR
Hangout presented by #PHPTestFestBrasil on 2017-11-14
https://www.youtube.com/watch?v=bBWkzZEadNo
Ensure that the "creating default object from empty value" warning is
always thrown. Previously some cases were missing the warning, in
particular those going through FETCH_OBJ_W rather than a dedicated
opcode (like ASSIGN_OBJ).
One slightly unfortunate side-effect of this change is that something
like $a->b->c = 'd' will now generate two warnings rather than one
when $a is null (one for property b, one for property c).
Contrary to the comments, these only hide constructors (old or new
style) if they a) are inherited b) come from a trait and c) are
aliased -- which doesn't make any sense at all.
As discussed in bug #77580, passing INVALID_HANDLE_VALUE for the
completion event improves compatibility with Wine/ReactOS. The timer
callback itself is supposed to complete fast enough, no behavior change
is to expect.
SO_USER_COOKIE primarly to be traced via dtrace instrumentation.
If SO_USER_COOKIE is available, the rest is so the former is
two years old while the rest is from around 2008.
Oniguruma does not consistently perform this validation itself (at least
on older versions), so make sure we check pattern encoding validity on the
PHP side.
A port of the original https://github.com/phpcommunity/phptestfest-php-src/pull/148 pull
request, created earlier during #PHPTestFest (User Group: PHP-WVL & PHPGent).
Expands the existing FTP delete command test with coverage for deletion of
non-existing files (which returns a 550 status code).
Rodrigo Prado de Jesus <royopa [at] gmail [dot] com>
User Group: PHPSP #PHPTestFestBrasil
Test function quotemeta() - using an empty string is given as str.
This test coverage line 2722 from file /ext/standard/string.c and is not a ZPP test.
http://gcov.php.net/PHP_HEAD/lcov_html/ext/standard/string.c.gcov.php#L2722
The run-tests script executes the cleaning section separately and interfers
with the running test itself less. This fixes two failing tests on Windows
platforms.
This patch follows previous license year ranges updates. With new
approach source code files now have simplified headers with license
information without year ranges.
This patch removes the so called local variables defined per
file basis for certain editors to properly show tab width, and
similar settings. These are mainly used by Vim and Emacs editors
yet with recent changes the once working definitions don't work
anymore in Vim without custom plugins or additional configuration.
Neither are these settings synced across the PHP code base.
A simpler and better approach is EditorConfig and fixing code
using some code style fixing tools in the future instead.
This patch also removes the so called modelines for Vim. Modelines
allow Vim editor specifically to set some editor configuration such as
syntax highlighting, indentation style and tab width to be set in the
first line or the last 5 lines per file basis. Since the php test
files have syntax highlighting already set in most editors properly and
EditorConfig takes care of the indentation settings, this patch removes
these as well for the Vim 6.0 and newer versions.
With the removal of local variables for certain editors such as
Emacs and Vim, the footer is also probably not needed anymore when
creating extensions using ext_skel.php script.
Additionally, Vim modelines for setting php syntax and some editor
settings has been removed from some *.phpt files. All these are
mostly not relevant for phpt files neither work properly in the
middle of the file.
I'm counting this towards the non-string needle deprecation from
https://wiki.php.net/rfc/deprecations_php_7_3. I wasn't aware that
mb_ereg_replace() is also affected by this issue. It's even more
ridiculous than usual here, because the integer is interpreted as
an ASCII codepoint, even though these are supposed to be multibyte
functions :(
This variable was dropped in the pkg-config migration, which resulted
in spurious warnings about using valgrind with external PCRE. Fix the
checks to use the right variable.
RFC: https://wiki.php.net/rfc/null_coalesce_equal_operator
$a ??= $b is $a ?? ($a = $b), with the difference that $a is only
evaluated once, to the degree that this is possible. In particular
in $a[foo()] ?? $b function foo() is only ever called once.
However, the variable access themselves will be reevaluated.
Formerly, a single option `--with-gd` was sufficient to enable the
extension, and to determine whether to use the system or the bundled
libgd depending on whether a directory was passed. Since pkg-config
determines the path automatically, we now offer `--enable-gd` (whether
the extension should be build) and `--with-external-gd` (whether to use
the system libgd).
Instead of interleaving creation of live-ranges with the main
compiler code, compute them in a separate pass over the opcodes
as part of pass_two. Additionally, do not keep live ranges
synchronized during optimization in opcache and instead use the
same mechanism to recompute them after optimization.
Even though libgd/libgd#492 is not a relevant bug fix for PHP, since
the binding doesn't use the `gdImage*Ptr()` functions at all, we're
porting the fix to stay in sync here.
get_gc is assumed to be non-NULL in master, and get_gc can be called
on the iterator wrapper if during generator GC, so we need to define
this handler. For now it's just a dummy, though for full support we'd
have to also add a get_gc iterator handler that is called here.
* --with-webp-dir becomes --with-webp
* --with-jpeg-dir becomes --with-jpeg
* --with-png-dir is removed. libpng is required.
* --with-zlib-dir is removed. zlib is required.
* --with-xpm-dir becomes --with-xpm.
We also enable --with-webp on Travis.
I thought these were redundant, because we already NULL out the
clone_obj object handler. However, it turns out that reflection is
using private __clone() to determine clonability (isCloneable) for
the case where we only have a class, rather than an object.
As such, removing these methods would be a BC break.
This reverts commit e7131a4e9f.
This reverts commit 55bd88ce0d.
openssl 0.9.8 in July 2005 first added pkg-config support, which is
earlier than the minimum supported version for php. This should
therefore be uiversally supported.
Previously this returned properties of a different object, including
INDIRECTs directly, which violates our invariants. Switch this to
only return properties for debugging purposes, without INDIRECTs.
If someone complains we can extend this to other purposes, as needed.
Since more than an year it not possible to create commands in the heap
but they are allocated on the stack and directly run. In this regard, it
doesn't make sense to have all the vararg stuff. Commands made sense
to be created and pushed onto a stack then a general executor will
run thru the stack and execute/handle the commands.
* PHP-7.3:
Fix bug #77418 - Heap overflow in utf32be_mbc_to_code
Add NEWS
[ci skip] Add NEWS
Fix more issues with encodilng length
Fix#77270: imagecolormatch Out Of Bounds Write on Heap
Fix bug #77380 (Global out of bounds read in xmlrpc base64 code)
Fix bug #77371 (heap buffer overflow in mb regex functions - compile_string_node)
Fix bug #77370 - check that we do not read past buffer end when parsing multibytes
Fix#77269: Potential unsigned underflow in gdImageScale
Fix bug #77247 (heap buffer overflow in phar_detect_phar_fname_ext)
Fix bug #77242 (heap out of bounds read in xmlrpc_decode())
Regenerate certs for openssl tests
* PHP-7.3:
Fix#77367: Negative size parameter in mb_split
Fix#77369 - memcpy with negative length via crafted DNS response
Fix more issues with encodilng length
Fix#77270: imagecolormatch Out Of Bounds Write on Heap
Fix bug #77380 (Global out of bounds read in xmlrpc base64 code)
Fix bug #77371 (heap buffer overflow in mb regex functions - compile_string_node)
Fix bug #77370 - check that we do not read past buffer end when parsing multibytes
Fix#77269: Potential unsigned underflow in gdImageScale
Fix bug #77247 (heap buffer overflow in phar_detect_phar_fname_ext)
Fix bug #77242 (heap out of bounds read in xmlrpc_decode())
Regenerate certs for openssl tests
Removed files used to provide building Zend engine separately using
Autoconf and Automake:
cd Zend && ./buildconf && ./configure
Since Zend engine is tightly integrated in the PHP code itself and
current root configure.ac and acinclude provide this patch removes
these. Removed files also included several obsolete macros and
Automake dependency which is not used in main *nix PHP build system.
We already detect the case where we're entirely outside a class --
now also check whether there actually is a parent.
This is a minor BC break, in that code that was never executed
might have previously contained an invalid parent:: reference without
generating an error.
Instead of representing this as a ZEND_AST_CLASS_CONST with a
"class" constant name.
Class constants and ::class are unrelated features that happen to
share syntax, so represent and handle them separately.
This parameter is always zero and not necessary to call pcre2_match.
I'm leaving the parameter behind on the _ex() variant, so the preg_flags
are still accessible in some way.
Use standard PKG_CHECK_MODULES macro and remove the fallback case.
The option is renamed to --with-external-pcre and no longer accepts
an explicit path -- it should be specified through pkg-config instead.
Albeit CSV is still a widespread data exchange format, it has never been
officially standardized. There exists, however, the “informational” RFC
4180[1] which has no notion of escape characters, but rather defines
`escaped` as strings enclosed in double-quotes where contained
double-quotes have to be doubled. While this concept is supported by
PHP's implementation (`$enclosure`), the `$escape` sometimes interferes,
so that `fgetcsv()` is unable to correctly parse externally generated
CSV, and `fputcsv()` is sometimes generating non-compliant CSV. Since
PHP's `$escape` concept is availble for many years, we cannot drop it
for BC reasons (even though many consider it as bug). Instead we allow
to pass an empty string as `$escape` parameter to the respective
functions, which results in ignoring/omitting any escaping, and as such
is more inline with RFC 4180. It is noteworthy that this is almost no
userland BC break, since formerly most functions did not accept an empty
string, and failed in this case. The only exception was `str_getcsv()`
which did accept an empty string, and used a backslash as escape
character then (which appears to be unintended behavior, anyway).
The changed functions are `fputcsv()`, `fgetcsv()` and `str_getcsv()`,
and also the `::setCsvControl()`, `::getCsvControl()`, `::fputcsv()`,
and `::fgetcsv()` methods of `SplFileObject`.
The implementation also changes the type of the escape parameter of the
PHP_APIs `php_fgetcsv()` and `php_fputcsv()` from `char` to `int`, where
`PHP_CSV_NO_ESCAPE` means to ignore/omit escaping. The parameter
accepts the same values as `isalpha()` and friends, i.e. “the value of
which shall be representable as an `unsigned char` or shall equal the
value of the macro `EOF`. If the argument has any other value, the
behavior is undefined.” This is a subtle BC break, since the character
`chr(128)` has the value `-1` if `char` is signed, and so likely would
be confused with `EOF` when converted to `int`. We consider this BC
break to be acceptable, since it's rather unlikely that anybody uses
`chr(128)` as escape character, and it easily can be fixed by casting
all `escape` arguments to `unsigned char`.
This patch implements the feature requests 38301[2] and 51496[3].
[1] <https://tools.ietf.org/html/rfc4180>
[2] <https://bugs.php.net/bug.php?id=38301>
[3] <https://bugs.php.net/bug.php?id=51496>
The `$mode` parameter of `imagecropauto()` defaults to `-1`. However,
`-1` is changed to `GD_CROP_DEFAULT` right away, so basically the
default is `GD_CROP_DEFAULT`, which is rather confusing and
unnecessary.
Therefore, we change the default to `IMG_CROP_DEFAULT`, but still allow
an explicit `-1` to be passed for BC reasons, in which case we trigger
a deprecation notice, so we can rid the `-1` support eventually.
The Autoconf macro AC_CONFIG_AUX_DIR can set the location of the
auxiliary build tools such as config.guess, config.sub, and bundled
libtool scripts and moves these bundled files from the root directory
to the build subdirectory.
Additionally some changes in this context or as a part of obsoletion:
- The LT_TARGETS variable in build/build2.mk file was once used as a part
of the Automake step. It's not used anymore and has been refactored to
separate makedist script directly.
- ltconfig is not used anymore since libtool 1.4+
cf8d1563c2
- phpize file locations for the config.guess, config.sub, and ltmain.sh
has been refactored accordingly.
curl 7.15.1 in December 2006 first added pkg-config support, which is
earlier than the minimum supported version for php. This should
therefore be uiversally supported.
While the major version is different, and the minor version may be
different, the revision and the extra should always be synced, as
recently discussed on the release manager's list. We add the
respective info to README.RELEASE_PROCESS, to not forget about it in
the future.
According to a respective RFC[1], the MD5 checksums have been removed
from the release notes and the API. Also, no checksums are generated
after uploading the tarballs to the user directories
on downloads.php.net.
[1] <https://wiki.php.net/rfc/release-md5-deprecation>
`GD_CROP_DEFAULT` and `GD_CROP_SIDES` are names of libgd constants, and
as such they are not relevant for userland developers. Therefore, we
replace them by the constant names of our wrapper, i.e.
`IMG_CROP_DEFAULT` and `IMG_CROP_SIDES`, respectively.
The AC_CHECK_FILE macro is designed to emit a warning about possible
cross compiling issues if file is not present on the target system.
Since the generated PHP lexer file is part of the build files and not
target system this can be simplified by a usual shell check with
`test -f` instead as in other usages.
Since upstream does not appear to move in any way[1], we sync our
behavior. Even though the BC break is ugly (which is the reason we
target master only), having to deal with different algorithms is even
worse for portable userland code.
[1] <https://github.com/libgd/libgd/issues/334>
Since cropping support has been added to our bundled libgd,
`gdImageAutoCrop` differs from upstream in that `GD_CROP_DEFAULT` falls
back on `GD_CROP_SIDES` if there is no transparent color in the image.
While this difference seem to be a useful improvement in our bundled
libgd, upstream has not yet signaled that there willing to back-port
it[1], so we revert it to stay in sync with upstream.
We also remove the additional NULL bailout at the end of the function,
which doesn't appear to be relevant any longer since bug 77198 has been
fixed.
[1] <https://github.com/libgd/libgd/issues/298>
While both source and destination buffers have the same size (6 bytes),
and this is unlikely to change in the future, we nonetheless fix the
illogical `strncpy` size.
Based on a pull request provided by Cristian Rodríguez.
`SQLite3::readOnly()` uses `sqlite3_stmt_readonly()` which is only
available as of libsqlite 3.7.4. For older SQLite3 versions we return
always `false`, which can be confusing. Instead of sticking with this
behavior, or even undefining the method for old SQLite3 versions, we
lift the requirements to SQLite 3.7.4 (released on 2010-12-08),
according to a respective discussion[1].
Since pdo_sqlite doesn't use `sqlite3_stmt_readonly()`, we stick with
the minimum requirement of SQLite 3.5.0.
[1] <https://github.com/php/php-src/pull/3614>
Since bug 77051 has been fixed, it is unlikely that any of the
`sqlite3_bind_*` calls will ever fail, but we add respective checks
nonetheless, and call `php_sqlite3_error()` in case of bind failures.
Serializing `SQLite3`, `SQLite3Stmt` and `SQLite3Result` instances is
possible but pointless, since unserializing results in uninitialized
instances, which will bail out of any method call. Therefore, we deny
serialization and unserialization in the first place.
* PHP-7.3:
Add DISPLAY_INI_ENTRIES for imap
Disable rsh/ssh functionality in imap by default (bug #77153)
Disable rsh/ssh functionality in imap by default (bug #77153)
FreeBSD provides MAP_ALIGNED_SUPER for a similar purpose as
MAP_HUGETLB on Linux, so add it as an alias.
Huge pages still have to be explicitly enabled through the
USE_ZEND_ALLOC_HUGE_PAGES environment variable.
Reflection classes already use NULLed clone_obj to signal that they
cannot be cloned, so it's not necessary to additionally declare a
throwing __clone() method.
Exceptions already prohibit cloning by setting clone_obj to NULL
(which is integrated with reflection). No need to additionally
define a dummy __clone() method.
To make installation experience better instead of only outputting
warning when re2c is not present this patch also exits if the PHP lexer
file(s) were not generated yet and re2c is not present on the system.
sapi/tests/ has been introduced as a generic means to test different
SAPIs[1]. run-tests2.php has later be renamed to server-tests.php and
recently been dropped[2]. However, the sapi/tests/test00?.php test
cases remained, even though they make no sense for run-tests.php,
since they use an unsupported format for the `--ENV--` section and the
completely unsupported `--HEADERS--` section, respectively. While
these tests ran successfully under run-tests.php, that was only by
accident, and they did not really test something useful. Therefore, we
remove these tests altogether.
sapi/tests/bug69487.phpt is actually a CGI test (CGI is enforced due to
the `--POST--` section), so we move it to sapi/cgi/tests, which leaves
sapi/tests/ empty. Thus, we also remove the sapi/ directory from
run-tests.php.
Finally, we remove the `--HEADERS--` and `--REQUEST--` sections from
the list of allowed run-tests.php sections.
[1] <http://git.php.net/?p=php-src.git;a=commit;h=b671380b6b5b6e1f4f235e810afa4199e989d2ba>
[2] <http://git.php.net/?p=php-src.git;a=commit;h=4f36acb9e65935aa657f1f22e2320a401bdbdad3>
With changes to (hopefully) correctly fall back if OpenSSL support
is missing. Furthermore the hard-coded dependency on ext/hash is
no longer an issue, as this extension is required in master.
This reverts commit 63072e9c0e, reversing
changes made to 4cbabb6852.
Do not create an array with two keys for every element. That's a
huge waste of memory. Instead we allocate a two-element structure.
This reduces memory usage and avoids having to perform hashtable
lookups to get at the data and priority.
The only thing this might impact negatively is the non-default
EXTR_BOTH mode, in which case the array has to be created anyway.
Besides the common `:param` notation to designate named parameters in
prepared statements, SQLite3 also supports `@param` and `$param`.
While the latter is mostly to support the Tcl programming language, and
would be confusing for PHP's sqlite3 binding due to the similarity with
string interpolation, the former is common under .NET and raises no
such issue. Therefore we add support for it.
This patch has been developed in cooperation with @BohwaZ.
This patch removes generated lexer files by re2c during the build
process so they are not tracked by Git yet can be still shipped when
PHP is released. The genfiles script additionally provides generation
of these lexer files when creating a release of the PHP source code.
The genfiles script refactorings:
- added file header
- echoing steps instead of comments
- cleaning only lines starting with `#line`
- eval removed in favor of direct executed commands
- the debug mode `set -x` removed
- script can be called from any path
- improved comments
This ensures tables remain legible when scrolling. If the user agent
does not support this feature, then it will gracefully degrade to
existing behaviour.
This patch implements: https://bugs.php.net/bug.php?id=77105
The parser files for ext/json are generated by bison from the *.y file.
Parser files in Zend and sapi/phpdbg already follow such approach of
these files being ignored from tracking in the Git repository and they
are shipped via the release packages later on. This way the end users
still don't need to have bison dependency installed to install PHP.
The genfiles script was refactored to generate the ext/json parser and
lexer files.
Certain sections, such as --EXTENSIONS--, --SKIPIF--, and --CLEAN--
require to be executed using only PHP CLI or PHP CGI SAPIs and cannot be
processed with the phpdbg.
Changes:
- Added introductory paragraph to explain the file for newcomers and
where to ignore local editor configurations as a common practice
when using Git.
- Patterns organized into several main context sections for a better
overview of the php-src directory structure and which files get
generated where.
- Added comments describing reasons for ignoring particular file(s).
- Patterns normalized according to Git ignore blob patterns. Absolute
paths are different than relative paths and similar fixes.
- Folders patterns have appended ending slash for readability to
distinguish them from files.
Removed ignore patterns:
- Removed `sapi/phpdbg/build` pattern since it doesn't seem to be
relevant anymore in the php-src repository.
- The `acconfig.h` pattern removed in favor of the more recent Autoconf
versions.
- Removed patterns for `.FBCIndex` `.FBCLockFolder` folders. These were
generated on old Mac OS systems [1] using the Sherlock [2] software
and can be today ignored using a global gitignore file if needed.
Newer macOS systems don't generate these anymore.
- Removed `php_version.h` pattern since the `main/php_version.h` file is
tracked by Git.
- Removed `*.mk` pattern since few *.mk files are tracked by Git and new
ones aren't generated during build process.
- Removed `**/tests/**/*.txt` pattern. This has been refactored via
595a395cb9 and current tests also clean
the generated `*.txt` files as soon as they are executed. Since there
are several `*.txt` files in the repository tracked by Git already and
renaming them would be counter productive so the pattern is now removed.
In case there will be a need to adjust it, we can rename the generated
`*.txt` into something else instead (ideally into `*.tmp`) to not
conflict with already tracked `*.txt` files in tests.
- Removed `ext/*/scan_makefile_in.awk` and `scan_makefile_in.awk` patterns
since these are already included in the `/ext/*/build/` pattern.
- Removed the `/ext/*/libs.mk` pattern since these auxilary Makefiles
have been used in previous build system and made obsolete via the commit
9d9d39a0de.
- Removed pattern for configuration header `ext/pdo_sqlite/sqlite3.h` since
it is not generated anymore in the current code.
- Removed `main/streams/build-defs.h` pattern since this file is not
generated.
- Removed `test.php3` pattern. These don't seem to be present in the current
PHP source code nor used since PHP 3.
- Removed `/ext/oci8/tests/*.vglog` pattern since the tests don't seem to
produce these files.
- Removed pattern for Visual Studio's `*.ncb` files. Since the Visual
Studio 2010 they are no longer used.
- Removed `ext/sqlite3/tests/phpsql*` pattern. It was part of the `ext/sqlite/tests`.
- Removed `shlibtool` pattern since the shlibtool file is not generated
anymore. It was once part of the libtool usage and customized naming
used in the previous build system in Makefiles. Such naming was later
removed and only libtool was used.
- Removed `meta_ccld` pattern since it is no longer relevant via
a5b55416f4
- Pattern `meta_cc` has been removed via e5176fe60a.
- Pattern `dynlib.m4` removed since the file is no longer generated via
9d9d39a0de
- Pattern `debug.log` removed since the Autoconf Macro PHP_DEBUG_MACRO
is not called anymore and this file not generated anymore.
- Pattern `results.txt` removed. It was once part of the logging results
of test files by Netware and previous Windows builds.
- Removed `_libs` pattern in favor of only `.libs`. These directories
were once generated by Automake and Libtool on systems that didn't
support the dot folder `.libs` (MS-DOS).
- Removed `*.opt` pattern since it was used by Visual Studio 6. Later
versions of Visual Studio don't use these anymore.
- Removed `*.plg` pattern. It was build log generated by Visual Studio 6.
- Removed `core` pattern since it was probably related to commit
fd7153b0f3 and is today not used.
- Removed ˙*.˙, ˙diff`, and ˙*.tgz` patterns since they don't seem to be
generated in the current code on the first glance. In case these will be
one day required to ignore again, they can be simply added back again.
- Removed Emacs specific interlock files pattern `.#*`. The number of
editors and IDEs specific files is very big. Emacs editors also produce
some other files such as `#*#` for autosave feature, which were previously
not ignored. To fully and properly support all additional patterns these
could be added to .gitignore files. A much more sustainable and better
practice is to use a global gitignore file on the local system or the
`.git/info/exclude` per repository basis for these in particular.
Added patterns:
- Added new pattern for `*.obj` standard object files in the Windows
section. These are generated by Visual Studio.
- Added `/ext/*/run-tests.php` patterns for extensions.
- Reversed patterns for particular tracked `/win32/build/Makefile`, `config.h`
files and `.patch` files.
- Added patterns for generated dtrace files `/ext/oci8/oci8_dtrace_gen.h`
Other changes:
- All `conftest*` patterns merged into a single one.
- Pattern `*.mem` made more specific where the memory check files get
generated by `run-tests.php -m`, i.e. `**/tests/**/*.mem`.
- Fixed `/Zend/zend_dtrace_gen.h.bak` `hbak` typo.
[1] http://www.westwind.com/reference/OS-X/invisibles.html
[2] https://en.wikipedia.org/wiki/Sherlock_(software)
The parser files for phpdbg are generated by bison from the *.y file.
Parser files in Zend already follows such approach of these files being
ignored from tracking in the Git repository and they are shipped via
the release packages. This way the end user still don't need to have
bison dependency installed to install PHP.
The genfiles script was refactored to generate the phpdbg parser and lexer
files.
Empty comment in phpdbg parser y template file has been changed to the
YACC compliant /* empty */ instead of custom one.
A dynamic property may be shadowed by a private/protected property.
Make sure we check property accessibility for non-indirect
properties as well.
Closes#3626.
The Automake and aclocal were part of the previous PHP build system
where Automake created Makefile from the Makefile.in templates and
aclocal was used to produce the aclocal.m4. They were removed as
dependencies via 9d9d39a0de and
e715fb00f8.
The `crlf` Git attribute has been long deprecated and made obsolete. A
replacement for using `-crlf` is `-text`. That way Git doesn't do
newlines normalization on any system.
Since the phpt files don't need specific line endings in any files
except for some special CR characters in some, this can be also removed
and line normalization can happen on all phpt files. The run-tests.php
file converts the CRLF to LF before running the test.
This patch also sets a `diff` attributes for all *.phpt files. It
helps showing diffs for some phpt files with special characters such
as ASCII control characters which makes Git recognize them as binary and
stops showing commit diffs otherwise
Some additional short info about used Git attributes has been also added
as comments for easier understanding.
Remove -text gitattribute
Current run-tests.php script produces the `*.php` files from the
*.phpt. So all *.php files in tests folders are ignored by Git.
To avoid confusion and to for bettere consistency this patch renames
two remaining tests/*/*.php files to *.inc and *.phar as current
practice in *.phpt files.
- The `ext/curl/tests/resonder/get.php` to .inc extension
- The `ext/phar/tests/files/pear2coverage.phar.php` to .phar extension
The `mkdep.awk` file was part of the previous *nix build system and was
used to create a .deps file with a list of dependencies that could be
processed by Automake further on.
Newer build system was done via 9d9d39a0de
and outdated files removed via 22815419f8
so the current file in the PHP source code is not used anymore.
Additionally, the *.slo files were processed by this file. The *.slo
files also used to be generated by older libtool so today, these don't
get generated anymore.
The `pear/scripts`, `pear/php-config`, `pear/phpize`, and
`pear/run-tests` used to be part of the PEAR installation. Now, the
pear installation PHAR file is directly downloaded from pear.php.net
instead.
The stamp-* files can be used as helpers for Makefiles to not redo
certain targets again. The stamp-h are mentioned in the Autoconf docs [1]
to help generate the config.h file.
Since the usage of stamp-h file was removed in 232afa4816
this patch cleans few obsolete occurrences.
This patch also removes two occurrences of `main/stamp-h1` and
`main/streams/stamp-h1` rules in the .gitignore file since they are not
generated with the current build systems anymore. The `stamp-h$am_indx` files
were once generated using aclocal and automake.
[1] https://www.gnu.org/software/autoconf/manual/autoconf-2.69/autoconf.html
The qa.php.net currently includes nice collection of information about
tests and how to run them. Instead of maintaining two locations of this
information, this patch removes the README.TESTING from the php-src
repo. Patch for qa.php.net has been sent separately to that repo.
The `makefile_am_files` was part of the previous build system where
automake was used to build Makefiles. Since 9d9d39a0de
this is not used anymore and can be removed.
The `bsd_converted` file was once used as a temporary locking mechanism
on BSD systems builds and has been made obsolete via commit
9d9d39a0de
so it can be also removed from the main .gitignore file.
The configuration-parser.c, configuration-parser.h,
configuration-parser.output and configuration-scanner.c were refactored
via 78194a47b7 and can be removed in the
.gitignore.
The buildconf.stamp file was used to store particular build time
information in the past and then got removed via the
6c6c0a630c
and the migration usage of the build/build.mk file only.
This patch adds missing newlines, trims multiple redundant final
newlines into a single one, and trims redundant leading newlines in all
*.phpt sections.
According to POSIX, a line is a sequence of zero or more non-' <newline>'
characters plus a terminating '<newline>' character. [1] Files should
normally have at least one final newline character.
C89 [2] and later standards [3] mention a final newline:
"A source file that is not empty shall end in a new-line character,
which shall not be immediately preceded by a backslash character."
Although it is not mandatory for all files to have a final newline
fixed, a more consistent and homogeneous approach brings less of commit
differences issues and a better development experience in certain text
editors and IDEs.
[1] http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_chap03.html#tag_03_206
[2] https://port70.net/~nsz/c/c89/c89-draft.html#2.1.1.2
[3] https://port70.net/~nsz/c/c99/n1256.html#5.1.1.2
* PHP-7.3:
Sync leading and final newlines in *.phpt sections
Sync leading and final newlines in *.phpt sections
Sync leading and final newlines in *.phpt sections
This patch adds missing newlines, trims multiple redundant final
newlines into a single one, and trims redundant leading newlines.
According to POSIX, a line is a sequence of zero or more non-' <newline>'
characters plus a terminating '<newline>' character. [1] Files should
normally have at least one final newline character.
C89 [2] and later standards [3] mention a final newline:
"A source file that is not empty shall end in a new-line character,
which shall not be immediately preceded by a backslash character."
Although it is not mandatory for all files to have a final newline
fixed, a more consistent and homogeneous approach brings less of commit
differences issues and a better development experience in certain text
editors and IDEs.
[1] http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_chap03.html#tag_03_206
[2] https://port70.net/~nsz/c/c89/c89-draft.html#2.1.1.2
[3] https://port70.net/~nsz/c/c99/n1256.html#5.1.1.2
* PHP-7.3:
Sync leading and final newlines in source code files
Sync leading and final newlines in source code files
Sync leading and final newlines in source code files
* PHP-7.3:
ext/sodium: sodium_pad(): do not copy any bytes if the string is empty
ext/sodium: Fix sodium_pad() with blocksize >= 256
ext/sodium: Use a correct max output size for base64 decoding
ext/sodium: Avoid shifts wider than 32 bits on size_t values
It is possible to pass flags when opening an SQLite database. For
Sqlite < 3.5.0 these are ignored, since `sqlite3_open` doesn't support
flags. Neither a warning or notice is raised in this case, nor is this
behavior documented in the PHP manual. Instead of fixing it either
way, we lift the requirement to SQLite 3.5.0 (released on 2007-09-04)
instead of the former SQLite 3.3.9 (released on 2007-01-04).
This patch simplifies line endings tracked in the Git repository and
syncs them to all include the LF style instead of the CRLF files.
Newline characters:
- LF (\n) (*nix and Mac)
- CRLF (\r\n) (Windows)
- CR (\r) (old Mac, obsolete)
To see which line endings are in the index and in the working copy the
following command can be used:
`git ls-files --eol`
Git additionally provides `.gitattributes` file to specify if some files
need to have specific line endings on all platforms (either CRLF or LF).
Changed files shouldn't cause issues on modern Windows platforms because
also Git can do output conversion is core.autocrlf=true is set on
Windows and use CRLF newlines in all files in the working tree.
Unless CRLF files are tracked specifically, Git by default tracks all
files in the index using LF newlines.
Instead of checking the type manually. Peculiarily the zpp
message does not use "array or object", but that's a general issue.
And I guess it's good that object support is not being advertised.
It turns out that all of these places ultimately only accept arrays,
not objects, so we can use Z_ARRVAL_P everywhere.
Also add _deref in a few places where the hash lookup result is
directly type-checked.
The Autoconf AC_PRESERVE_HELP_ORDER macro has been available since
Autoconf 2.59c [1] and in PHP it has been called conditionally on two places
to support older Autoconf versions. With recent updates and the macro
can be called unconditionally.
[1] http://git.savannah.gnu.org/cgit/autoconf.git/tree/NEWS
Instead overload get_properties_for for a few specific cases such
as array casts. This resolves the issue where ArrayObject
get_properties may violate engine invariants in some cases.
This resolves the long-standing issue where var_dump
a DateTime (etc) object makes a number of additional properties
accessible, which may also change other behaviors as a side-effect.
ICU 59 already requires C++11 by default. The minimum version required
by the core is 50, which is compiled with at least C++11 in many distros
as package defs tell. Headers for ICU versions between ICU 50 and 58 look
fine when included for C++11 compilation, the linking is thereof not affected.
The macro PHP_CXX_COMPILE_STDCXX is based on
https://www.gnu.org/software/autoconf-archive/ax_cxx_compile_stdcxx.html
The patch consists on allowing to read the corresponding switch in a user
defined variable instead of enforcing CXXFLAGS globally. That way, every
ext or SAPI can decide, which C++ standard is to be used. The
documentation is provided in the m4 file.
C++11 is already somewhat older standard, C++14 were better. However
issues with GCC < 5.0 and some other compilers are possibly to hit back.
Still there's some time to check for C++14 for ext/intl, too. Having said
that, C++11 in ext/intl and a mechanism to determine features is a good step
towards better C++ support.
Since there is no need to patch libsqlite3 for our purposes, and since
libsqlite3 ≥ 3.3.9 (which is our current requirement) is widely
available on distros, there is no reason anymore to bundle the library.
Besides removing the bundled libsqlite, and adapting the configuration
respectively, we also fix the use of the SQLITE_ENABLE_COLUMN_METADATA
compile time constant to detect whether sqlite3_column_table_name() is
available by a working feature detection (otherwise bug_42589.phpt
would fail). We also skip bug73068.phpt for libsqlite 3.11.0 to
3.14.1 which have a bug (<https://sqlite.org/src/info/ef360601>).
We also completely drop support for the obscure pdo_sqlite_external
extension (which could have been enabled on Windows only by passing
`--pdo-sqlite-external` to configure), since it is not needed anymore.
Furthermore, we remove references to the bundled libsqlite from
Makefile.gcov, CONTRIBUTING.md and README.REDIST.BINS.
run-tests.php assumes the contents of an ENV section are a simple list
of key-value-pairs, instead of evaluating the section contents as PHP
script which server-tests.php did. Since server-tests.php is gone, we
remove the superfluous and confusing scripting.
* 'master' of git.php.net:php-src:
Second round of fixing compiler warnings in ext/hash. Lots of the signatures that previously was unsigned int is now size_t, there was a fair bit of inconsistency there already and this commit should make it all sync nicely
Write to correct smart_str
Revert "Fix compiler warnings in ext/hash"
Instead of ZEND_HASH_FOREACH. As a side-effect, this fixes a latent
bug in _addmethod, where a zval was interpreted as a zval*.
Also apply some optimizations to getProperties() while at it: For
declared properties, use the HT key instead of unmangling the
property name. For dynamic properties check INDIRECT instead of
looking up prop info to determine if the property is dynamic.
* 'master' of git.php.net:php-src: (29 commits)
Fix the deplister rule to not ignore the .c file (Anatol)
Update .gitignore to include the Windows deplister program (win32/build/deplister.c)
Bug > Feature Request
NEWS and UPGRADING
Fixed bug #75479
Fix test
Fix some tests and improve coverage for Windows in SPL
Use already set variable
Fix reflection arguments for sodium_memzero function
Deprecate unbinding of $this of non-static methods
Generalize compile_typename
Fixed bug #76737
Fixed bug #72635
Remove and refactor ext/spl/examples
Remove outdated soap examples
Remove unused ext/bz2/php_bz2.def
Remove redundant ce from reflection property_reference
Only store zend_type inside reflection type_reference
Fixed bug #76946
Bump versions for 7.1.24-dev
...
* Letargie-feature-getcolumnmeta:
Removing last unused
Last few changes : . force the nls_date_format . add the scale to the return of the function . add tests on some function return . removing unused variables
cs
Changes : . Add the distinction between NUMBER and FLOAT types . Changing BFLOAT text to be BINARY_FLOAT . Changing BDOUBLE text to be BINARY_DOUBLE . Add the data types names for NCHAR, NVARCHAR and NCLOB . Few changes in the tests
fixing comments
Update oci_statement.c
Add the PDOStatement::getColumnMeta() function to the pdo_oci driver
Merge branch 'feature-getcolumnmeta' of https://github.com/Letargie/php-src into Letargie-feature-getcolumnmeta
* 'feature-getcolumnmeta' of https://github.com/Letargie/php-src:
Removing last unused
Last few changes : . force the nls_date_format . add the scale to the return of the function . add tests on some function return . removing unused variables
cs
Changes : . Add the distinction between NUMBER and FLOAT types . Changing BFLOAT text to be BINARY_FLOAT . Changing BDOUBLE text to be BINARY_DOUBLE . Add the data types names for NCHAR, NVARCHAR and NCLOB . Few changes in the tests
fixing comments
Update oci_statement.c
Add the PDOStatement::getColumnMeta() function to the pdo_oci driver
Static calls to non-static methods have been fully deprecated in
PHP 7.0 as part of https://wiki.php.net/rfc/reclassify_e_strict.
A combination of ReflectionMethod::getClosure() ("fake closures")
and Closure::bindTo() etc can be used to achieve the same behavior.
This commit ensures that a deprecation notice will be thrown also
in this case.
- Test file from ext/spl/examples has been moved to ext/spl/tests
- Other custom SPL examples and implementations were removed in favor of
the PHP manual.
We don't need the full arg_info and the baggage associated with it.
Storing the type should be safe, as it's part of structures which
are never released during a request (arg_info and in the future
prop_info).
If a property access would normally result in a magic method call,
but the property is subject to an active recursion guard, the
access should behave as if the magic method does not exist.
This commit fixes one instance where this was not the case -- we
should have been generating a property access error, but instead
the operation simply did not do anything.
This can avoid some confusions with .gitignored autogenerated `.sh`
files in test directories and to distinguish the autogenerated shell
files from the tracked in Git ones more easily.
The ctype.xml file has been added with the early ctype extension
implementation and hasn't been updated with the recent changes. Also
file is not used anymore.
- ext/simplexml/tests/bug25756_1.xml
- ext/simplexml/tests/bug25756_2.xml
- ext/simplexml/tests/bug25756.xsd
Added via 503d74aa29 and then removed via
9e29f17493
- ext/mysqli/tests/cacert.pem
- ext/mysqli/tests/client-cert.pem
- ext/mysqli/tests/client-key.pem
Added via e9f9f66f2e and then removed via
6d51b7b2e3
- ext/gd/tests/simpletext私はガラスを食べられます.jpg
Added via 3d3f11ede4 and never used
- ext/gd/tests/src.png
Added via cc938b5df0 and never used
- ext/zlib/tests/gzgetss.test
Added via d536ecac5c as a file content of the
ext/zlib/tests/gzgetss.gzbut but never used in tests directly. Removed for
better clarity of zlib tests
- ext/soap/tests/interop/Round3/GroupD/round3_groupD_import2_absolute.wsdl
- ext/soap/tests/interop/Round4/GroupG/round4_groupG_mimerpc.wsdl
- ext/soap/tests/interop/Round4/GroupG/round4_groupG_mimedoc.wsdl
Added via 1d25fc5c7b and never used
- ext/reflection/tests/exception.inc
Removed via 9f8ba2e8a1
- ext/phar/tests/files/extracted.inc
Removed via 549bf83bd1
- ext/phar/tests/cache_list/files/extracted.inc
Added via 05c3104097 and never used
When including files in PHP tests a good practice so far has been to use
the *.inc extension. This patch renames few *.p5c files that are
included in phpt files.
This patch adds a php-src directory structure introduction to the
contributing guidelines. Thanks to the discussion at [1] and PHP
manual.
[1] https://github.com/php/php-src/pull/3520
The `<time.h>` header file is part of the standard C89 headers [1] and
on current systems can be included unconditionally.
Since PHP requires at least C89 or greater, the `HAVE_TIME_H` symbol
defined by Autoconf in ext/pdo_sqlite/config.m4 [2] can be ommitted and
simplifed.
Additionally, since PHP didn't define `HAVE_TIME_H` prior in the
configure.ac the occurrence of this symbol in cli can be removed.
Refs:
[1] https://port70.net/~nsz/c/c89/c89-draft.html#4.1.2
[2] https://git.savannah.gnu.org/cgit/autoconf.git/tree/lib/autoconf/headers.m4
. Add the distinction between NUMBER and FLOAT types
. Changing BFLOAT text to be BINARY_FLOAT
. Changing BDOUBLE text to be BINARY_DOUBLE
. Add the data types names for NCHAR, NVARCHAR and NCLOB
. Few changes in the tests
Also, my commit is correcting a bug in the function oci_stmt_describe. PDO override the values given for the precision using the content of the pdo_column_data property 'precision'.
That value was set in oci_stmt_describe but with the value of OCI_ATTR_SCALE, causing a bug for the precision value returned by getColumnMeta() (values between 12800 and 14000 probably due to the conversion of a sb1 to a zend_ulong)
As far as I know server-tests.php is unmaintained and no longer
used for anything nowadays. Some of the new sections are supported
by run-tests.php itself nowadays and some server-based tests
(using the CLI server or php-fpm) are performed through different
means.
Dropping server-tests.php has been discussed in a bunch of PR
discussions already, so let's go ahead with it...
The C89 standard and later defines the `<string.h>` header as part of
the standard headers [1] and on current systems it is always present.
Code included also `<strings.h>` header as an alterinative in some
files. This kind of check was relevant on some older systems where the
`<strings.h>` file included definitions for the C89 compliant
`<string.h>`. Today such alternative check is not required anymore. The
`<strings.h>` file is part of the POSIX definition these days.
Also Autoconf suggests doing this and relying on C89 or above [2] and [3].
This patch also cleans few unused `<strings.h>` inclusions in the libmbfl.
[1]: https://port70.net/~nsz/c/c89/c89-draft.html#4.1.2
[2]: http://git.savannah.gnu.org/cgit/autoconf.git/tree/lib/autoconf/headers.m4
[3]: https://www.gnu.org/software/autoconf/manual/autoconf-2.69/autoconf.html
Files `config.guess` and `config.sub` are committed in the php-src
repository so these don't need to be ignored for the root directory.
For the extensions these can be ignored in case of building extension
inside the php-src with phpize.
This is an internal glibc macro, it should not be necessary to use
it if we already define _GNU_SOURCE (we do through
AC_USE_SYSTEM_EXTENSIONS). Needing to use __USE_GNU generally
indicates an inclusion order problem (libc header included before
config.h).
Newer MySQL versions are stricter about invalid values. Three issues
are fixed:
* Don't use negative values with ZEROFILL. ZEROFILL implies UNSIGNED.
* Use a legal TIMESTAMP value. TIMESTAMP does not accept a Unix timestamp.
* Specify BIGINT values as strings, to avoid overflows.
Support for libt1 has been removed from ext/gd in PHP 7.0 by
7dcfdbbee4, so we no longer need
to install this package. It is also no longer available on newer
Ubuntu versions.
HAVE_STDDEF_H guards have been removed by
6c1ff61a36, because the header is
always available in C89. This removes a __GNUC__ guard for the
same header.
The FormatMessage API needs to LocalFree the delivered error messages.
In cases where messages are delivered in non ASCII compatible encoding,
the messages might be unreadable. This aligns the error message encoding
with the encoding settings in PHP, the focus is UTF-8 as default.
Initialize error buffer
Avoid code duplication
free_object_storage intentionally does not free the object allocations
to make sure that they show up as leaks. However, if the object is in
the GC root buffer, the later GC run may end up freeing the allocation,
hiding the leak.
Avoid this by moving the final GC run before free_object_storage, the
way it was done before fast_shutdown was integrated into core.
This patch makes few remaining PHP development tools files
executable and adds a shebang to them.
The `#!/usr/bin/env php` shebang provides running the script via
`./script.php` and uses env to find PHP script location on the system.
At the same time it still provides running the script with a user
defined PHP location using `php script.php`. Shebang is not visible in
the output of the generated file.
Some INI processors allow to specify empty values by just giving the
key without the equals sign, for instance MySQL and Python. It appears
to be sensible to add this possibility to our INI parser, so that it
can be used for such INI files as well. We choose NULL as the value of
empty values.
This syntactical enhancement is a (minor) BC break, though, as can be
seen by the necessary change to bug49692.ini. The “comment” formerly
has been simply ignored, but now it would be parsed as key with an
empty value.
This PR is based on Adam's former patch.
According to the “Deprecate and remove INTL_IDNA_VARIANT_2003” RFC[1],
we change the default of the $variant parameter of `idn_to_ascii()` and
`idn_to_utf8()` from `INTL_IDNA_VARIANT_2003` to
`INTL_IDNA_VARIANT_UTS46`.
[1] <https://wiki.php.net/rfc/deprecate-and-remove-intl_idna_variant_2003>
Given that ICU is a set of lively developed libraries, that ICU 50.1
has been released on 2012-11-05, and PHP 7.4 is scheduled to be
released seven years after it, we consider it appropriate to ditch
these legacy versions.
Particularly, that would be a reasonable groundwork to implement part
two of the “Deprecate and remove INTL_IDNA_VARIANT_2003” RFC[1], namely
to default idn_to_ascii()'s and idn_to_utf8()'s $variant parameter to
INTL_IDNA_VARIANT_UTS46, which is not defined in ICU < 4.6.
See also the related discussion on internals@[2].
[1] <https://wiki.php.net/rfc/deprecate-and-remove-intl_idna_variant_2003>
[2] <http://news.php.net/php.internals/101626>ff
Header `<errno.h>` is part of the standard C89 headers [1] and on
current systems checking is not need anymore since PHP requires at
least C89. This is noted also by Autoconf itself in the docs and
code [2].
The Autoconf check defined the `HAVE_ERRNO_H` symbol when building PHP
with sockets extension or fpm sapi. This symbol is not utilized across
the PHP source code except in the current version of bundled GD library
which has worked ok so far also with sockets extension or fpm sapi
disabled anyway.
Refs:
[1] https://port70.net/~nsz/c/c89/c89-draft.html#4.1.2
[2] https://git.savannah.gnu.org/cgit/autoconf.git/tree/lib/autoconf/headers.m4
The `<assert.h>` header file is part of the standard C89 headers [1] and
on older systems there needed to be also a manual check if header is
present.
Since PHP requires at least C89 manual check and the `HAVE_ASSERT_H`
symbol defined by Autoconf in configure.ac can be both removed [2].
This patch also removes unused <assert.h> includes where c files don't
use the `assert()` macro.
Refs:
[1] https://port70.net/~nsz/c/c89/c89-draft.html#4.2
[2] https://git.savannah.gnu.org/cgit/autoconf.git/tree/lib/autoconf/headers.m4
Formerly, we sent output regarding memory leaks always to the debugger
on Windows, but this appears to be not useful especially for the PHPTs,
which usually are not run under a debugger, and so important info will
not be available there.
The `<stddef.h>` header file is part of the standard C89 headers [1] and
on current systems there is no need for a manual check if header is
present.
Since PHP requires at least C89 the `HAVE_STDDEF_H` symbol isn't defined
by Autoconf anywhere else anymore [2] and accross the PHP source code
the header is included unconditionally already.
This patch syncs this also for the bundled libmbfl which is maintaned as
a fork in php-src.
Refs:
[1] https://port70.net/~nsz/c/c89/c89-draft.html#4.1.2
[2] https://git.savannah.gnu.org/cgit/autoconf.git/tree/lib/autoconf/headers.m4
Autoconf 2.59d (released in 2006) [1] started promoting several macros
as not relevant for newer systems, including the AC_FUNC_UTIME_NULL.
This macro checks if `utime(file, NULL)` sets file's timestamp to the
current time and defines the `HAVE_UTIME_NULL` symbol. This check was
relevant on very old systems (for example, 4.3BSD released in 1986) and
today can be omitted for systems with utime since it should be well
supported by now. [2]
Refs:
[1] http://git.savannah.gnu.org/cgit/autoconf.git/tree/NEWS
[2] https://www.gnu.org/software/autoconf/manual/autoconf-2.69/autoconf.html
Autoconf 2.59d (released in 2006) [1] started promoting several macros
as not relevant for newer systems anymore, including the `AC_FUNC_MEMCMP`.
On some old systems such as SunOS 4.1.3 (EOL in 2003) and NeXT x86
OpenStep (discontinued) the `memcmp` function wasn't present or it
didn't work properly. [2]
On current systems including at least Solaris 10+ this check is not
relevant anymore.
Refs:
[1] http://git.savannah.gnu.org/cgit/autoconf.git/tree/NEWS
[2] https://www.gnu.org/software/autoconf/manual/autoconf-2.69/autoconf.html
Autoconf 2.59d (released in 2006) [1] started promoting several macros
as not relevant for newer systems anymore, including the `AC_HEADER_DIRENT`.
This macro checks which header defines the `DIR` type. If `<dirent.h>`
is available it defines the `HAVE_DIRENT_H` symbol. Since the `<dirent.h>`
header is already checked in the `configure.ac`, this check is not needed
anymore. This macro also additionally checks for SCO Xenix (discontinued,
latest release 1989) dir and x libraries. [2]
Commit 6ed790685f introduced also
`<sys/dir.h>`. This header exists from times of UNIX System V and
provided definition of DIR type on these systems such as 4.3BSD.
Today `<sys/dir.h>` is kept for backwards compatibility and includes
the `<dirent.h>` on current systems. With `dirent.h>` present this
include is no longer required.
Refs:
[1] http://git.savannah.gnu.org/cgit/autoconf.git/tree/NEWS
[2] https://www.gnu.org/software/autoconf/manual/autoconf-2.69/autoconf.html
Remove unused dirent.h includes
On line 512, I clarified the sentence which uses effect, previously the sentence was awkward.
On line 656, the wrong word is used, I changed affect to effect.
Autoconf 2.59d (released in 2006) [1] started promoting several macros
as not relevant for newer systems anymore, including the `AC_HEADER_TIME`.
This macro checks if both `<sys/time.h>` and `<time.h>` can be included
at the same time and defines the `TIME_WITH_SYS_TIME` and
`HAVE_SYS_TIME_H` symbols. On current system such check is not relevant
anymore because in case both headers are present both can be also
included at the same time.
This patch simplifies this checking.
Refs:
[1] http://git.savannah.gnu.org/cgit/autoconf.git/tree/NEWS
[2] https://www.gnu.org/software/autoconf/manual/autoconf-2.69/autoconf.html
Autoconf 2.59d (released in 2006) [1] started promoting several macros
as not relevant for newer systems, including the `AC_C_CONST`.
The `const` keyword is used in C since C89. On old systems some compilers
lacked the `const` and this macro defined it to be empty. This check was
relevant on systems with compilers before C89 and on current systems it
can be omitted. [2]
PHP also requires at least C89 so `const` is always available.
Refs:
[1] http://git.savannah.gnu.org/cgit/autoconf.git/tree/NEWS
[2] https://www.gnu.org/software/autoconf/manual/autoconf-2.69/autoconf.html
This reverts changes made in 2f700384 and 4fcaebba. This test is
supposed to run
only on Windows indeed, as in a Unix environment with a mail server
configured,
an email will be send. On Windows, only emptying the SMTP ini directives
is enough.
Autoconf 2.59d (released in 2006) 1 started promoting several macros
as not relevant for newer systems anymore, including the AC_FUNC_VPRINTF.
This macro checks for presence of the vprint function otherwise checks
for presence of the _doprnt function. This check was relevant on very
old systems and today can be omitted since it should be well supported
by now. [2]
Also PHP doesn't use the HAVE_VPRINTF or HAVE_DOPRNT symbols.
Refs:
[1] http://git.savannah.gnu.org/cgit/autoconf.git/tree/NEWS
[2] https://www.gnu.org/software/autoconf/manual/autoconf-2.69/autoconf.html
Autoconf 2.59d (released in 2006) [1] started promoting several macros
as not relevant for newer systems anymore, including the `AC_STRUCT_TM`.
This macro checks if `struct tm` is defined in either `<sys/time.h>` or
`<time.h>` and defines the `TM_IN_SYS_TIME` symbol accordingly. This
check was relevant in times before the C89 for some embedded systems,
microcontrollers or very old systems. For newer systems it can be
avoided and the `<time.h>` should be included instead since current
systems should be well supported by now. [2]
Since PHP requires at least C89, this patch removes the obsolescent call
and time.h checks.
Refs:
- [1]: http://git.savannah.gnu.org/cgit/autoconf.git/tree/NEWS
- [2]: https://www.gnu.org/software/autoconf/manual/autoconf-2.69/html_node/Particular-Structures.html
This patch makes several scripts and PHP development tools files
executable and adds more proper shebangs to the PHP scripts.
The `#!/usr/bin/env php` shebang provides running the script via
`./script.php` and uses env to find PHP script location on the system.
At the same time it still provides running the script with a user
defined PHP location using `php script.php`.
The AC_ARG_PROGRAM Autoconf macro provides program name transformations
when installing. This patch implements #64517 and prepares the
implementation for the request #60518.
In ./configure --help it additionally outputs --program-prefix=PREFIX,
--program-suffix=SUFFIX and the upcoming --program-transform-name=PROGRAM
option.
Macro AC_ARG_PROGRAM is available since Autoconf 2.0 and needs to be
called after the AC_CANONICAL_TARGET macro.
Refs:
- https://www.gnu.org/software/autoconf/manual/autoconf-2.69/html_node/Transforming-Names.html
- move relevant parts into win32
- general cleanup
- use Windows API and fallback to POSIX
- improve filetime to timestamp conversion
- improve stat/fsat
- handle ino by using file index
- handle st_dev by using volume serial number
The inode implementation is based on file indexes from NTFS. On 32-bit,
fake inodes are shown, that may lead to unexpeted results. 64-bit
implementation is most reliable.
Documentation for the SPL was once rendered at
http://www.php.net/~helly/php/ext/spl/
With upgrades of the PHP manual, the SPL documentation moved there
instead and PHP files for the SPL docs in the PHP source code directory
were outdated and removed via 9003d8a4cf
Since Autoconf 2.50+ macro AC_STRUCT_ST_BLOCKS defines the new
HAVE_STRUCT_STAT_ST_BLOCKS symbol and has deprecated the previous
HAVE_ST_BLOCKS.
PHP 5.3 required Autoconf 2.13 (released in 1999) or newer, since PHP
5.4 the autoconf 2.59 (released in 2003) or newer was required, and
since PHP 7.2, autoconf 2.64 (released in 2008) or newer is required.
Attempting to create a symlink on Windows via the CLI SAPI without
administrative rights is likely to fail with ERROR_PRIVILEGE_NOT_HELD.
Therefore we map this Windows error to EACCES to get better diagnostics
and to avoid the assert() failure in debug mode.
These macros are supposed to behave like POSIX's symlink() and link(),
respectively, on POSIX compliant systems and on Windows.
Future scope: merge link.c and link_win32.c
This attribute is a boolean value. It is taken from the return value of
sqlite3_stmt_readonly(), indicating if and only if the prepared statement makes
no direct changes to the content of the database.
Travis logs are usually inspected after the build completes, by which time
progress info is useless.
Signed-off-by: Philip Prindeville <philipp@redfish-solutions.com>
-@files=`find . -name \*.gcda -o -name \*.gcno -o -name \*.da -o -name \*.c -o -name \*.h | sed -e 's/^\.\///'| sed -e 's/\.gcda//g' -e 's/\.gcno//g' -e 's/\.da//g'|$(EGREP)$(LCOV_INCLUDE)| sed -e 's/.libs/zzzz/g'| sort | sed -e 's/zzzz/.libs/g'| uniq`;\
for x in $$files;do\
echo -n . ;\
y=`echo$$x | sed -e 's!\.libs/!!'`;\
dir=lcov_data/`dirname $$x`;\
test -d "$$dir"||$(mkinstalldirs)"$$dir";\
iftest -f "$(top_srcdir)/$$y.c";then\
ln -f -s $(top_srcdir)/$$y.c lcov_data/$$y.c;\
fi;\
iftest -f "$(top_srcdir)/$$y.h";then\
ln -f -s $(top_srcdir)/$$y.h lcov_data/$$y.h;\
fi;\
iftest -f "$(top_srcdir)/$$y.re";then\
ln -f -s $(top_srcdir)/$$y.re lcov_data/$$y.re;\
fi;\
iftest -f "$(top_srcdir)/$$y.y";then\
ln -f -s $(top_srcdir)/$$y.y lcov_data/$$y.y;\
fi;\
iftest -f "$(top_srcdir)/$$y.l";then\
ln -f -s $(top_srcdir)/$$y.l lcov_data/$$y.l;\
fi;\
iftest -f "$(top_srcdir)/$$y";then\
ln -f -s $(top_srcdir)/$$y lcov_data/$$y;\
fi;\
iftest -f "$(top_builddir)/$$y.c";then\
ln -f -s $(top_builddir)/$$y.c lcov_data/$$y.c;\
fi;\
iftest -f "$$x.gcno";then\
cp $$x.gcno lcov_data/$$y.gcno ;\
fi;\
iftest -f "$$x.gcda";then\
cp $$x.gcda lcov_data/$$y.gcda ;\
fi;\
iftest -f "$$x.da";then\
cp $$x.da lcov_data/$$y.da ;\
fi;\
iftest -f "$$x.bb";then\
cp $$x.bb lcov_data/$$y.bb ;\
fi;\
iftest -f "$$x.bbg";then\
cp $$x.bbg lcov_data/$$y.bbg ;\
fi;\
done;\
for dir in ext/bcmath/libbcmath ext/date/lib ext/fileinfo/libmagic ext/gd/libgd ext/mbstring/libmbfl ext/mbstring/oniguruma ext/pcre/pcrelib ext/pdo_sqlite/libsqlite ext/sqlite3/libsqlite ext/xmlrpc/libxmlrpc ext/zip/lib;do\
if test $bison_version_num -ge $bison_version_min; then
php_cv_bison_version="$bison_version (ok)"
for bison_check_version in $bison_version_exclude; do
if test "$bison_version" = "$bison_check_version"; then
php_cv_bison_version=invalid
break
fi
done
fi
fi
])
fi
case $php_cv_bison_version in
""|invalid[)]
bison_msg="This bison version is not supported for regeneration of the Zend/PHP parsers (found: $bison_version, min: $bison_version_min, excluded: $bison_version_exclude)."
AC_MSG_WARN([$bison_msg])
YACC="exit 0;"
;;
esac
])
AC_DEFUN([ZEND_FP_EXCEPT],[
AC_CACHE_CHECK(whether fp_except is defined, ac_cv_type_fp_except,[
function my_error_handler($errno, $errstr, $errfile, $errline) {
var_dump($errstr);
}
set_error_handler('my_error_handler');
class NoneTest
{
function __toString() {
@@ -44,7 +38,11 @@ class NoneTest
}
$o = new NoneTest;
echo $o;
try {
echo $o;
} catch (Error $e) {
echo $e->getMessage(), "\n";
}
echo "===THROW===\n";
@@ -58,17 +56,16 @@ class ErrorTest
$o = new ErrorTest;
try {
echo $o;
}
catch (Exception $e) {
echo "Got the exception\n";
} catch (Exception $e) {
echo $e->getMessage(), "\n";
}
?>
===DONE===
--EXPECTF--
--EXPECT--
Hello World!
===NONE===
string(%d) "Method NoneTest::__toString() must return a string value"
Method NoneTest::__toString() must return a string value
===THROW===
Fatal error: Method ErrorTest::__toString() must not throw an exception, caught Exception: This is an error! in %sbug26166.php on line %d
This is an error!
===DONE===
Loaded 100 of 8122 files, more files were not shown because too many files have changed in this diff.
Show more
Reference in new issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Derick Rethans