On Windows, closing a file which is locked may not immediately remove
the lock. The `LockFileEx()` documentation states:
| Therefore, it is recommended that your process explicitly unlock all
| files it has locked when it terminates.
We comply, and also use the macro `LOCK_EX` instead of the magic number
`2`.
Closes GH-8925.
We revert the commits which caused this regression from the PHP-8.0 and
PHP-8.1 branches for now. We keep it in "master" because of PR #8833
which may offer a proper fix without BC break.
When casting a `variant` to `int`, we need to heed the proper `zval`
type, which is an signed 64bit integer on x64, while `VT_INT` is only
a signed 32bit integer.
Closes GH-8779.
A connection string may contain just a single key, but
PHP used ";" as the heuristic to detect if a string was a connection
string versus plain DSN. However, a single-key connection string
would get treated like a DSN name, i.e. "DSN=*LOCAL". This makes it
so that "=" is used, as a connection string must contain a key.
Closes GH-8748.
It introduces a single function to check file paths passed to OpenSSL
functions. It expands the path, check null bytes and finally does
an open basedir check.
when passing an int to a string enum. Previously, the int was coerced to
a string. The JIT skips parameter clean up when unnecessary. In this
particular case, passing int to from(int|string) normally doesn't cause
a coercion so no dtor for the $value zval is generated.
To circumvent this we avoid coersion by explicitly allowing ints and
converting them to strings ourselves. Then we can free it appropriately.
See GH-8518
Closes GH-8633
Indirect Branch Tracking (IBT) is part of Intel's Control-Flow
Enforcement Technology (CET). IBT is hardware based, forward edge
Control-Flow-Integrity mechanism where any indirect CALL/JMP must target
an ENDBR instruction or suffer #CP.
This commit adds IBT support for fiber:
1. Add endbr32/64 in assembly
2. Inform compiler jump_fcontext may return via indirect branch
Furthermore:
gcc support CET since v8.1 and set it to default since gcc 11. That is,
the ELF header of sapi/cli/php has a property named IBT. However, such
property is lost since PHP8.1 because the assembly introduced by Fiber.
This commit also fixes this.
Closes GH-8339
Signed-off-by: Chen, Hu <hu1.chen@intel.com>
Co-authored-by: Christoph M. Becker <cmbecker69@gmx.de>
When stripping the namespace prefix, we can assume that this does not
contain any colons, while the rest of the name may contain colons.
Hence we must not use `strrchr()` but rather `strchr()` instead.
Closes GH-8543.
Extensions may (and do) write to stderr in mshutdown and similar. In
the best case, with the stderr stream closed, it's just swallowed.
However, some libraries will do things like try to detect color, and
these will outright fail and cause an error path to be taken.
Extensions may (and do) write to stderr in mshutdown and similar. In
the best case, with the stderr stream closed, it's just swallowed.
However, some libraries will do things like try to detect color, and
these will outright fail and cause an error path to be taken.
The unexpected EOF failure was introduced in OpenSSL 3.0 to prevent
truncation attack. However there are many non complaint servers and
it is causing break for many users including potential majority
of those where the truncation attack is not applicable. For that reason
we try to keep behavior consitent with older OpenSSL versions which is
also the path chosen by some other languages and web servers.
Closes GH-8369
This issue might happen if there is change of the fcgi stream when
the buffer is full. Then the empty record is created which signals
end of stream which is incorrect.
The actual fix without a test was contributed by GitHub user @loveharmful
in GH-3198.
We need to prevent integer overflow to eventually stop the iteration.
A test case doesn't appear sensible for this, because even on 32bit
architectures a respective test easily runs for a few minutes.
Closes GH-8447.
The fix for GH-7953 introduced a regression by being to deliberate
adding the respective headers. These must only be added, if the
handler starts, but is not finalizing.
Closes GH-8353.
We need to deref any references passed in the `$values` array. While
we could handle this in the type switch, doing it right away in the
foreach loop makes that more explicit, and also circumvents the missing
range checks for integers which are not passed as int or double.
Closes GH-8407.
This change introduces subsequent kill of the process when idle process quit
(SIGQUIT) does not succeed. It can happen in some situations and means that FPM
is not able to scale down in dynamic pm. Using SIGKILL fixes the issue.
In 7502c86342, I adjusted the number of error markers emitted on
invalid UTF-8 text to be more consistent with mbstring's behavior on
other text encodings (generally, it emits one error marker for one
unexpected byte). I didn't expect that anybody would actually care one
way or the other, but felt that it was better to be consistent than
not.
Later, Martin Auswöger kindly pointed out that the WHATWG encoding
specification, which governs how various text encodings are handled
by web browsers, does actually specify how many error markers should
be generated for any given piece of invalid UTF-8 text.
Until now, we have never really paid much attention to the WHATWG
specification, but we do want to comply with as many relevant
specifications as possible. And since PHP is commonly used for web
applications, compatibility with the behavior of web browsers is
obviously a good thing.
The fix introduces early locking of scoreboard when it is updated
which prevents the race condition causing an incorrect number of
active processes being set.
Instead of using the unsupported `%I64u` and `%I64d` format specifiers
on Windows, we use the portable `PRIu64` and `PRId64` specifiers.
The `L64()` macro and the `my_longlong` typedef should be adapted as
well, as the `i64` literal suffix is still supported by MSVC, but using
`LL` or `ll` is recommended[1], and the standard `int64_t` is available
there anyway. This is not urgent, though.
[1] <https://docs.microsoft.com/en-us/cpp/cpp/numeric-boolean-and-pointer-literals-cpp?view=msvc-170#integer-literals>
Closes GH-8268.
We must not allow to serialize already finalized `HashContext`s, since
the internal context is already freed. Since there is not much point
in serializing finalized `HashContext`s, we just bail out in that case.
Closes GH-8265.
Apparently, this has been forgotten when PHP 8.0.17RC1 and 8.0.18RC1
had been tagged.
We also fix the version of the fix for GH-8253, which didn't make it
into PHP 8.0.18RC1.
Integer parameters are stored in `zend_long` values, which have 64 bits
on LLP64, but `long` has only 32 bits there.
Adding a test might be overkill, because the broken behavior could
already be observed when running pg_select_001.phpt on Windows debug
builds, which report the stack corruption.
Closes GH-8263.
When fetching into objects, we need to create object style hash tables,
i.e. where numeric column names are stored as string keys instead of
integer keys. Instead of the slightly more efficient alternative to
create the desired hash table in the first place, we go for the more
readable implementation and convert the array style hash table using
`zend_symtable_to_proptable()`.
Co-authored-by: Kamil Tekiela <tekiela246@gmail.com>
Closes GH-8189.
First, we must not free the current line before we call
`spl_filesystem_file_read_csv()`, because then the `current_line` will
not be properly updated. Since the EOF check is superfluous here, we
move that part of the code to the branch for subtypes. This issue has
been introduced by the fix for bug 75917.
Second, we only must increase the `current_line` if we're not reading
ahead. This issue has been introduced by the fix for bug 62004.
Closes GH-8138.
==109253== 280 (56 direct, 224 indirect) bytes in 1 blocks are definitely lost in loss record 4 of 4
==109253== at 0x483B7F3: malloc (in /usr/lib/x86_64-linux-gnu/valgrind/vgpreload_memcheck-amd64-linux.so)
==109253== by 0x6D9FA2: __zend_malloc (zend_alloc.c:3068)
==109253== by 0x745138: zend_add_attribute (zend_attributes.c:226)
==109253== by 0x6680D1: zend_add_parameter_attribute (zend_attributes.h:102)
==109253== by 0x66B787: zm_startup_zend_test (test.c:478)
==109253== by 0x7224CD: zend_startup_module_ex (zend_API.c:2202)
==109253== by 0x72252C: zend_startup_module_zval (zend_API.c:2217)
==109253== by 0x734288: zend_hash_apply (zend_hash.c:2011)
==109253== by 0x722C30: zend_startup_modules (zend_API.c:2328)
==109253== by 0x67409B: php_module_startup (main.c:2256)
==109253== by 0x88EDDE: php_cli_startup (php_cli.c:409)
==109253== by 0x890F61: main (php_cli.c:1334)
This is achieved by tracking the observers on the run_time_cache (with a fixed amount of slots, 2 for each observer).
That way round, if the run_time_cache is freed all associated observer data is as well.
This approach has been chosen, as to avoid any ABI or API breakage.
Future versions may for example choose to provide a hookable API for run_time_cache freeing or similar.
When we need to evaluate constant ASTs, we always have to do that in
the scope where the constant has been defined, which may be a parent
of the `ReflectionClass`'s scope.
Closes GH-8106.
If either the first or second operand of `range()` may be a string, we
must not exclude the possibility that the result may be an array of
longs.
Closes GH-8131.
fpm_scoreboard_copy locks the scoreboard while copying the scoreboard and all
proc scoreboards. proc scoreboards are locked one by one while copying each
struct. The old implementation (inside fpm_handle_status_request) only briefly
locked the scoreboard while copying the scorebard.
Closes GH-7931
Co-authored-by: Jakub Zelenka <bukka@php.net>