]> granicus.if.org Git - php/log
php
5 years agoDisable RTLD_DEEPBIND under memory sanitizer
Nikita Popov [Fri, 28 Jun 2019 16:15:38 +0000 (18:15 +0200)]
Disable RTLD_DEEPBIND under memory sanitizer

This was already done for address sanitizer.

5 years agoUpdate NEWS [ci skip]
Peter Kokot [Mon, 1 Jul 2019 01:10:09 +0000 (03:10 +0200)]
Update NEWS [ci skip]

5 years agoMerge branch 'PHP-7.3' into PHP-7.4
Peter Kokot [Mon, 1 Jul 2019 00:55:02 +0000 (02:55 +0200)]
Merge branch 'PHP-7.3' into PHP-7.4

* PHP-7.3:
  Update NEWS [ci skip]

5 years agoUpdate NEWS [ci skip]
Peter Kokot [Mon, 1 Jul 2019 00:54:42 +0000 (02:54 +0200)]
Update NEWS [ci skip]

5 years agoMerge branch 'PHP-7.3' into PHP-7.4
Peter Kokot [Mon, 1 Jul 2019 00:53:10 +0000 (02:53 +0200)]
Merge branch 'PHP-7.3' into PHP-7.4

* PHP-7.3:
  Fix bug #78197: PCRE2 version check

5 years agoFix bug #78197: PCRE2 version check
Peter Kokot [Sun, 23 Jun 2019 13:46:47 +0000 (15:46 +0200)]
Fix bug #78197: PCRE2 version check

When external PCRE library version is of format ##.##-RC# the check
failed.

5 years agoUpdate changelogs [ci skip]
Peter Kokot [Sun, 30 Jun 2019 23:35:29 +0000 (01:35 +0200)]
Update changelogs [ci skip]

5 years agoSync functions checks
Peter Kokot [Fri, 28 Jun 2019 21:10:37 +0000 (23:10 +0200)]
Sync functions checks

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
  560ed89bfbb56526251e9bd2fbcb27e5932641d8 which uses PHP's own implementation)
- strpncpy check removed, added via a8c9e893b6911c785f698d724b2ff61e7aebb737 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

5 years agoRemove unused build specific symbols
Peter Kokot [Wed, 26 Jun 2019 19:55:26 +0000 (21:55 +0200)]
Remove unused build specific symbols

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 f35af61ef8fd10f2d1bada1909e888293fb519b8.

- PHP_OCI8_SHARED_LIBADD and PHP_OCI8_DIR were removed via
  c33f88ccb08ca05ebf822ef94cb3fa41a1838226

And other are not used anywhere across the php-src codebase

5 years agoFix phar tests with hash extension
Peter Kokot [Sat, 29 Jun 2019 12:40:31 +0000 (14:40 +0200)]
Fix phar tests with hash extension

- The hash extension is now always available.

- Remove unfinished test
  The Phar::PGP currently doesn't exist yet.

5 years agoFix missing file and remove SKIP section
Peter Kokot [Sat, 29 Jun 2019 12:46:51 +0000 (14:46 +0200)]
Fix missing file and remove SKIP section

Test now passes on VC 14, 15 and 16 with current MSVC runtimes.

5 years agoRemove HAVE_LIMITS_H check
Peter Kokot [Sun, 30 Jun 2019 21:44:45 +0000 (23:44 +0200)]
Remove HAVE_LIMITS_H check

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
aae5907cb7e775f16bedf61b010b8692c12a2843

PHP extensions out there shouldn't rely on symbols defined during the
build anyway and neither they do on this particular symbol anymore.

5 years agoRemove HAVE_DEV_ARANDOM
Peter Kokot [Sun, 30 Jun 2019 02:38:07 +0000 (04:38 +0200)]
Remove HAVE_DEV_ARANDOM

The arandom check is not needed anymore since the implementations of:
3467526a65bfb15eaf9ec49a0b5673b84e26bca4
and
6554f721f770c99037f07d465a7d610568576ce4

5 years agoMerge branch 'PHP-7.3' into PHP-7.4
Derick Rethans [Sun, 30 Jun 2019 14:50:55 +0000 (15:50 +0100)]
Merge branch 'PHP-7.3' into PHP-7.4

5 years agoFixed tests due to changed timezone data
Derick Rethans [Sun, 30 Jun 2019 14:50:38 +0000 (15:50 +0100)]
Fixed tests due to changed timezone data

5 years agoUpdate timelib to 2018.02
Derick Rethans [Sun, 30 Jun 2019 14:45:13 +0000 (15:45 +0100)]
Update timelib to 2018.02

5 years agoMerge branch 'PHP-7.3' into PHP-7.4
Anatol Belski [Sat, 29 Jun 2019 15:41:29 +0000 (17:41 +0200)]
Merge branch 'PHP-7.3' into PHP-7.4

* PHP-7.3:
  Update NEWS [ci skip]

5 years agoUpdate NEWS [ci skip]
Anatol Belski [Sat, 29 Jun 2019 15:40:58 +0000 (17:40 +0200)]
Update NEWS [ci skip]

5 years agoMerge branch 'PHP-7.2' into PHP-7.3
Anatol Belski [Sat, 29 Jun 2019 15:40:14 +0000 (17:40 +0200)]
Merge branch 'PHP-7.2' into PHP-7.3

* PHP-7.2:
  Update NEWS [ci skip]

5 years agoUpdate NEWS [ci skip]
Anatol Belski [Sat, 29 Jun 2019 15:39:45 +0000 (17:39 +0200)]
Update NEWS [ci skip]

5 years agoUpdate libmagic.patch [ci skip]
Anatol Belski [Sat, 29 Jun 2019 15:35:25 +0000 (17:35 +0200)]
Update libmagic.patch [ci skip]

5 years agoSet buffer to NULL to prevent double free (Kamil Dudka at redhat)
Anatol Belski [Sat, 29 Jun 2019 15:10:29 +0000 (17:10 +0200)]
Set buffer to NULL to prevent double free (Kamil Dudka at redhat)

Based on the upstream patch d13204e8da43f1d8a98cf3f74591cde02ecdb2f7.

5 years agoFix magic data
Anatol Belski [Sat, 29 Jun 2019 14:51:41 +0000 (16:51 +0200)]
Fix magic data

5 years agoFix memory leak in COM ctor
Christoph M. Becker [Sat, 29 Jun 2019 12:47:06 +0000 (14:47 +0200)]
Fix memory leak in COM ctor

We have to free strings returned from php_com_olestring_to_string().

5 years agoMerge branch 'PHP-7.3' into PHP-7.4
Anatol Belski [Sat, 29 Jun 2019 11:31:22 +0000 (13:31 +0200)]
Merge branch 'PHP-7.3' into PHP-7.4

* PHP-7.3:
  Fix typo
  Add tests from fix #78183 from PHP-7.2
  Fix #78183: finfo_file shows wrong mime-type for .tga file

5 years agoFix typo
Anatol Belski [Sat, 29 Jun 2019 11:30:35 +0000 (13:30 +0200)]
Fix typo

5 years agoAdd tests from fix #78183 from PHP-7.2
Joshua Westerheide [Sat, 29 Jun 2019 09:59:08 +0000 (11:59 +0200)]
Add tests from fix #78183 from PHP-7.2

5 years agoMerge branch 'PHP-7.2' into PHP-7.3
Anatol Belski [Sat, 29 Jun 2019 11:29:41 +0000 (13:29 +0200)]
Merge branch 'PHP-7.2' into PHP-7.3

* PHP-7.2:
  Fix #78183: finfo_file shows wrong mime-type for .tga file

5 years agoFix #78183: finfo_file shows wrong mime-type for .tga file
Joshua Westerheide [Fri, 28 Jun 2019 13:07:14 +0000 (15:07 +0200)]
Fix #78183: finfo_file shows wrong mime-type for .tga file

Due to a bug in the underlying libmagic 5.31, .tga images returned mime type "image/x-tgaimage/x-tga".

5 years agoimplement support for workers in cli-server on platforms supporting fork
Joe Watkins [Thu, 27 Jun 2019 13:09:58 +0000 (15:09 +0200)]
implement support for workers in cli-server on platforms supporting fork

5 years agoadd ini config details to php-config
Joe Watkins [Wed, 26 Jun 2019 10:49:01 +0000 (12:49 +0200)]
add ini config details to php-config

5 years agoClean up files after the tests are run
Peter Kokot [Thu, 27 Jun 2019 00:21:31 +0000 (02:21 +0200)]
Clean up files after the tests are run

5 years agoUpdate changelogs [ci skip]
Peter Kokot [Fri, 28 Jun 2019 18:14:22 +0000 (20:14 +0200)]
Update changelogs [ci skip]

5 years agoRemove HAVE_ERRNO_H from libgd
Peter Kokot [Thu, 27 Jun 2019 22:58:26 +0000 (00:58 +0200)]
Remove HAVE_ERRNO_H from libgd

The check for errno.h has been removed via
50b9ef8d9435d23be40cea7fb484a02e99fdb617

Upstream libgd library is also patched via
https://github.com/libgd/libgd/commit/1e7f93922fb3adf9f131d7e94aa13386062ffe11

5 years agoUse strcmp() in phpdbg_eol_global_update
Nikita Popov [Fri, 28 Jun 2019 15:24:30 +0000 (17:24 +0200)]
Use strcmp() in phpdbg_eol_global_update

memcmp() only makes sense on equal length strings, and here we don't
know anything about the length of the input.

5 years agoDisable ifunc resolvers if memory sanitizer is used
Nikita Popov [Fri, 28 Jun 2019 15:15:56 +0000 (17:15 +0200)]
Disable ifunc resolvers if memory sanitizer is used

Just marking them as no_sanitize("memory") is unforunately not
sufficient, as the function still gets instrumented -- the attribute
only disables reporting.

5 years agoRemove phpdbg_xml_asprintf()
Nikita Popov [Fri, 28 Jun 2019 14:59:52 +0000 (16:59 +0200)]
Remove phpdbg_xml_asprintf()

Clang doesn't like the use of promoted zend_bool in va_start. As
the function isn't used anyway, I'm simply dropping it...

5 years agoMark ifunc resolver functions as unused
Nikita Popov [Fri, 28 Jun 2019 14:55:54 +0000 (16:55 +0200)]
Mark ifunc resolver functions as unused

It looks like clang doesn't see that these are used through
__attribute__((ifunc)) and generates a warning.

5 years agoFix arginfo leak in disable_functions
Nikita Popov [Fri, 28 Jun 2019 14:36:58 +0000 (16:36 +0200)]
Fix arginfo leak in disable_functions

Arginfo is allocated if types are used, we need to free it.

5 years agoMerge branch 'PHP-7.3' into PHP-7.4
Nikita Popov [Fri, 28 Jun 2019 14:09:23 +0000 (16:09 +0200)]
Merge branch 'PHP-7.3' into PHP-7.4

5 years agoMerge branch 'PHP-7.2' into PHP-7.3
Nikita Popov [Fri, 28 Jun 2019 14:09:09 +0000 (16:09 +0200)]
Merge branch 'PHP-7.2' into PHP-7.3

5 years agoFix bugs in AST printer
sunnyeo [Fri, 28 Jun 2019 14:06:01 +0000 (16:06 +0200)]
Fix bugs in AST printer

Closes GH-4324.

5 years agoSet up asan+ubsan scheduled build on azure
Nikita Popov [Thu, 20 Jun 2019 14:51:14 +0000 (16:51 +0200)]
Set up asan+ubsan scheduled build on azure

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.

5 years agoFix out of bounds read in sccp
Nikita Popov [Fri, 28 Jun 2019 10:38:28 +0000 (12:38 +0200)]
Fix out of bounds read in sccp

5 years agoAvoid reliance on arena details on phpdbg oplog
Nikita Popov [Fri, 28 Jun 2019 10:32:54 +0000 (12:32 +0200)]
Avoid reliance on arena details on phpdbg oplog

Instead of guessing what the address of the first arena allocation
is going to be, embed the sentinel in the oplog_list structure
directly.

5 years agoUpdate error message in libmysql test
Nikita Popov [Fri, 28 Jun 2019 12:36:10 +0000 (14:36 +0200)]
Update error message in libmysql test

5 years agoFix compilation errors when building against libmysql
Sebastian Pop [Fri, 21 Jun 2019 07:15:35 +0000 (07:15 +0000)]
Fix compilation errors when building against libmysql

Closes GH-4316.

5 years agoFix out-of-bounds read in array compilation
Nikita Popov [Fri, 28 Jun 2019 08:24:56 +0000 (10:24 +0200)]
Fix out-of-bounds read in array compilation

UNPACK elements only have one child. Don't access the second one
until we have excluded this case.

5 years agoAdd tracked arena allocator
Nikita Popov [Fri, 28 Jun 2019 08:19:18 +0000 (10:19 +0200)]
Add tracked arena allocator

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.

5 years agoMerge branch 'PHP-7.3' into PHP-7.4
Christoph M. Becker [Fri, 28 Jun 2019 10:37:28 +0000 (12:37 +0200)]
Merge branch 'PHP-7.3' into PHP-7.4

* PHP-7.3:
  Fix bug #78192 PDO SQLite SegFault when reuse statement after schema has changed

5 years agoMerge branch 'PHP-7.2' into PHP-7.3
Christoph M. Becker [Fri, 28 Jun 2019 10:36:33 +0000 (12:36 +0200)]
Merge branch 'PHP-7.2' into PHP-7.3

* PHP-7.2:
  Fix bug #78192 PDO SQLite SegFault when reuse statement after schema has changed

5 years agoFix bug #78192 PDO SQLite SegFault when reuse statement after schema has changed
Vincent [Wed, 26 Jun 2019 09:37:08 +0000 (11:37 +0200)]
Fix bug #78192 PDO SQLite SegFault when reuse statement after schema has changed

Reset stmt->columns when column count changed on new execution of prepared statement

5 years agoAdd support for proc_open() with a command array
Nikita Popov [Mon, 24 Jun 2019 10:53:40 +0000 (12:53 +0200)]
Add support for proc_open() with a command array

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).

5 years agoReduce overhead
Dmitry Stogov [Fri, 28 Jun 2019 09:03:01 +0000 (12:03 +0300)]
Reduce overhead

5 years agoimplement tsrm_is_shutdown API
Joe Watkins [Wed, 26 Jun 2019 11:18:50 +0000 (13:18 +0200)]
implement tsrm_is_shutdown API

5 years agoUse the portable zend_strndup() instead of strndup()
Christoph M. Becker [Fri, 28 Jun 2019 08:20:50 +0000 (10:20 +0200)]
Use the portable zend_strndup() instead of strndup()

5 years agoFixed use-after-free
Dmitry Stogov [Fri, 28 Jun 2019 07:43:56 +0000 (10:43 +0300)]
Fixed use-after-free

5 years agoFix mysqlnd memory leak
Nikita Popov [Thu, 27 Jun 2019 14:52:19 +0000 (16:52 +0200)]
Fix mysqlnd memory leak

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.

5 years agoFree ZTS lock in opcache
Nikita Popov [Thu, 27 Jun 2019 14:01:13 +0000 (16:01 +0200)]
Free ZTS lock in opcache

For some reason this only shows up as a leak when using phpdbg.

5 years agoFix init_file leak in phpdbg
Nikita Popov [Thu, 27 Jun 2019 13:43:50 +0000 (15:43 +0200)]
Fix init_file leak in phpdbg

5 years agorun-tests: Don't pass --INI-- settings to --SKIPIF--
Nikita Popov [Thu, 27 Jun 2019 12:07:20 +0000 (14:07 +0200)]
run-tests: Don't pass --INI-- settings to --SKIPIF--

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.

5 years agoFix custom heap free
Nikita Popov [Fri, 28 Jun 2019 07:13:45 +0000 (09:13 +0200)]
Fix custom heap free

This seems to be designed around the use-case where the custom
allocator is a wrapper around ZMM.

5 years agoBugfix #78208 Needs rehash with an unknown algo should always return true.
Sara Golemon [Thu, 27 Jun 2019 23:00:32 +0000 (19:00 -0400)]
Bugfix #78208 Needs rehash with an unknown algo should always return true.

5 years agoCatch up with some recent changes [ci skip]
Peter Kokot [Tue, 25 Jun 2019 20:32:54 +0000 (22:32 +0200)]
Catch up with some recent changes [ci skip]

- Mention mysqli::stat()
- Mention removal of HAVE_* for always defined extensions
- Mention removed symbols defined during the build

5 years agoRemove HAVE_STRCOLL check
Peter Kokot [Sun, 23 Jun 2019 13:51:24 +0000 (15:51 +0200)]
Remove HAVE_STRCOLL check

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

5 years agoSync HAVE_HASH, HAVE_HASH_EXT, PHAR_HASH_OK symbols
Peter Kokot [Wed, 26 Jun 2019 01:01:45 +0000 (03:01 +0200)]
Sync HAVE_HASH, HAVE_HASH_EXT, PHAR_HASH_OK symbols

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.

5 years agoRemove posix_getlogin_basic.phpt test
Nikita Popov [Thu, 27 Jun 2019 14:15:13 +0000 (16:15 +0200)]
Remove posix_getlogin_basic.phpt test

The premise of this test (that the getlogin user has to be the same
as the geteuid user) is incorrect.

5 years agoFix opcache blacklist leak
Nikita Popov [Thu, 27 Jun 2019 11:09:14 +0000 (13:09 +0200)]
Fix opcache blacklist leak

Don't recompute (and leak) the regex for every glob element, only
do this once at the end.

5 years agoFix phpdbg shutdown order
Nikita Popov [Thu, 27 Jun 2019 10:46:06 +0000 (12:46 +0200)]
Fix phpdbg shutdown order

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.

5 years agoDon't use zmm for PHPDBG_G(exec)
Nikita Popov [Thu, 27 Jun 2019 10:41:41 +0000 (12:41 +0200)]
Don't use zmm for PHPDBG_G(exec)

Uses system allocator when backing up settings ... let's avoid
unnecessary confusion.

5 years agoFix persistent XML memory leaks in SOAP
Nikita Popov [Thu, 27 Jun 2019 08:57:49 +0000 (10:57 +0200)]
Fix persistent XML memory leaks in SOAP

SOAP uses a horrible bailout based error handling approach -- avoid
leaking persistent XML memory by catching bailouts in a number of
places.

5 years agoAdd tracked allocator mode
Nikita Popov [Thu, 27 Jun 2019 08:30:45 +0000 (10:30 +0200)]
Add tracked allocator mode

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.

5 years agoShow actual enchant version if possible
Christoph M. Becker [Thu, 27 Jun 2019 11:04:49 +0000 (13:04 +0200)]
Show actual enchant version if possible

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/>

5 years agoSeparate common code abd eliminate useless checks
Dmitry Stogov [Thu, 27 Jun 2019 10:11:08 +0000 (13:11 +0300)]
Separate common code abd eliminate useless checks

5 years agoInline hot path
Dmitry Stogov [Thu, 27 Jun 2019 08:27:53 +0000 (11:27 +0300)]
Inline hot path

5 years agoFix persistent connection leak in mysqli
Nikita Popov [Wed, 26 Jun 2019 14:51:12 +0000 (16:51 +0200)]
Fix persistent connection leak in mysqli

5 years agoFix use-after-free in stream freeing during shutdown
Nikita Popov [Wed, 26 Jun 2019 08:58:29 +0000 (10:58 +0200)]
Fix use-after-free in stream freeing during shutdown

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!)

5 years agoClean headers checks
Peter Kokot [Sun, 23 Jun 2019 20:04:03 +0000 (22:04 +0200)]
Clean headers checks

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.

5 years agoAvoid double checks on early binding
Dmitry Stogov [Wed, 26 Jun 2019 23:13:06 +0000 (02:13 +0300)]
Avoid double checks on early binding

5 years agoRemove warnings for bison and re2c checks
Peter Kokot [Sun, 23 Jun 2019 13:53:19 +0000 (15:53 +0200)]
Remove warnings for bison and re2c 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

5 years agoFix concurrent testing of bug 61964
Peter Kokot [Tue, 25 Jun 2019 19:23:07 +0000 (21:23 +0200)]
Fix concurrent testing of bug 61964

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

5 years agoFix posix_errno test without PCNTL dep
Peter Kokot [Tue, 25 Jun 2019 18:21:16 +0000 (20:21 +0200)]
Fix posix_errno test without PCNTL dep

When PCNTL extension is not enabled, the SIGKILL constant is also not
available.

5 years agoRemove unused HAVE_DATE and HAVE_SPL symbols
Peter Kokot [Wed, 26 Jun 2019 00:36:44 +0000 (02:36 +0200)]
Remove unused HAVE_DATE and HAVE_SPL symbols

These two are never defined anymore since PHP 7.4 and can be
considered to have always available date and hash extensions.

5 years agoRemove HAVE_STRTOD and strtod check
Peter Kokot [Sun, 23 Jun 2019 13:41:43 +0000 (15:41 +0200)]
Remove HAVE_STRTOD and strtod check

There is no need to check for the strtod function. It is part of C89
standard which PHP-7.4+ supports.

http://port70.net/~nsz/c/c89/c89-draft.html#4.10.1.4

5 years agoSet Computer English language for credits sorting
Peter Kokot [Sun, 23 Jun 2019 13:43:25 +0000 (15:43 +0200)]
Set Computer English language for credits sorting

This makes the script a bit more portable when used on systems with
different LC_ALL and LANG settings.

5 years agoMerge branch 'PHP-7.3' into PHP-7.4
Christoph M. Becker [Wed, 26 Jun 2019 15:02:08 +0000 (17:02 +0200)]
Merge branch 'PHP-7.3' into PHP-7.4

* PHP-7.3:
  Remove stream_socket_sendto.phpt

5 years agoMerge branch 'PHP-7.2' into PHP-7.3
Christoph M. Becker [Wed, 26 Jun 2019 15:01:29 +0000 (17:01 +0200)]
Merge branch 'PHP-7.2' into PHP-7.3

* PHP-7.2:
  Remove stream_socket_sendto.phpt

5 years agoRemove stream_socket_sendto.phpt
Nikita Popov [Wed, 26 Jun 2019 14:58:37 +0000 (16:58 +0200)]
Remove stream_socket_sendto.phpt

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.

(cherry picked from commit 84333cad674890174c47f3c21b1b1cf85a4276ac)

5 years agoPrivate methods don't have to be duplicated
Dmitry Stogov [Wed, 26 Jun 2019 10:27:13 +0000 (13:27 +0300)]
Private methods don't have to be duplicated

5 years agoFix xml doc leak in soap
Nikita Popov [Tue, 25 Jun 2019 14:57:09 +0000 (16:57 +0200)]
Fix xml doc leak in soap

5 years agoUse ldap_destroy instead of ldap_unbind_ext
Nikita Popov [Tue, 25 Jun 2019 13:49:28 +0000 (15:49 +0200)]
Use ldap_destroy instead of ldap_unbind_ext

5 years agoFix fpm limit_extensions leak
Nikita Popov [Tue, 25 Jun 2019 13:01:58 +0000 (15:01 +0200)]
Fix fpm limit_extensions leak

The logic here is somewhat ugly because the worker pool structure
gets freed early, so we need to explicitly transfer ownership.

5 years agoReorder conditions to minimize number of checks on fast path
Dmitry Stogov [Tue, 25 Jun 2019 22:00:31 +0000 (01:00 +0300)]
Reorder conditions to minimize number of checks on fast path

5 years agoRemove always true/false conditions, remove dead conde and simplify code.
Dmitry Stogov [Tue, 25 Jun 2019 21:32:22 +0000 (00:32 +0300)]
Remove always true/false conditions, remove dead conde and simplify code.

5 years agoReplace previosly checked conditions by ZEND_ASSERT()
Dmitry Stogov [Tue, 25 Jun 2019 15:26:56 +0000 (18:26 +0300)]
Replace previosly checked conditions by ZEND_ASSERT()

5 years agoFixed variance check for abstract constructor during erlay binding
Dmitry Stogov [Tue, 25 Jun 2019 14:43:46 +0000 (17:43 +0300)]
Fixed variance check for abstract constructor during erlay binding

5 years agoCleanup
Dmitry Stogov [Tue, 25 Jun 2019 13:37:42 +0000 (16:37 +0300)]
Cleanup

5 years agoFFI: Perform bitfield operations byte-wise
Nikita Popov [Tue, 25 Jun 2019 09:32:18 +0000 (11:32 +0200)]
FFI: Perform bitfield operations byte-wise

Otherwise we may perform reads/writes outside the allocation, as
already happens in 032.phpt.

5 years agoFix out of bounds write in phpdbg
Nikita Popov [Tue, 25 Jun 2019 11:08:28 +0000 (13:08 +0200)]
Fix out of bounds write in phpdbg

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.

5 years agoFix CURLINFO_COOKIELIST leak
Nikita Popov [Tue, 25 Jun 2019 11:17:39 +0000 (13:17 +0200)]
Fix CURLINFO_COOKIELIST leak