cURL 8.11.0 added a couple of packages to `Requires.private`, but these
packages are irrelevant when building against a shared libcurl. For
some reason, these private requirements are checked when we're doing
`pkg-config --cflags` (that happens with the preinstalled pkg-config
0.29.2, as well as with pkgconf 2.3.0). To avoid further messing with
these packages, we just drop the `Requires.private` line from
libcurl.pc.
See GH-16741 for more details.
Closes GH-16783.
Reproducer: https://github.com/php/php-src/issues/16727#issuecomment-2466256317
The root cause is a data race between two different threads:
1) We allocate a lower cased name for an anonymous class here:
f97353f228/Zend/zend_compile.c (L8109)
2) This gets looked up as an interned string here:
f97353f228/Zend/zend_compile.c (L8112)
Assuming that there are uppercase symbols in the string and therefore
`lcname != name` and that `lcname` is not yet in the interned string table,
the pointer value of `lcname` won't change.
3) Here we add the string into the interned string table:
f97353f228/Zend/zend_compile.c (L8223)
However, in the meantime another thread could've added the string into the interned string table.
This means that the following code will run, indirectly called via the `LITERAL_STR` macro,
freeing `lcname`: 62e53e6f49/ext/opcache/ZendAccelerator.c (L572-L575)
4) In the reproducer we then access the freed `lcname` string here:
f97353f228/Zend/zend_compile.c (L8229)
This is solved in my patch by retrieving the interned string pointer
and putting it in `lcname`.
Closes GH-16748.
There are two issues:
1) There's an off-by-one in the check for the minimum file size for a
tar (i.e. `>` instead of `>=`).
2) The loop in the tar parsing parses a header, and then unconditionally
reads the next one. However, that doesn't necessarily exist.
Instead, we remove the loop condition and check for the end of the
file before reading the next header. Note that we can't use
php_stream_eof as the flag may not be set yet when we're already at
the end.
Closes GH-16700.
Adding two exit early safeguards in the *nix configuration build script:
1) Given the initial cd into the build tree fails (the project root),
the `buildconf` script exits with non-zero status (failure).
2) Given the grep command does not exist or `configure.ac` AC_INIT [1]
expectations are unmet, the buildconf script exits non-zero.
Additionally quoting the pathname to cd into and the empty CD_PATH
parameter for portability, also for systems that are using a
non-portable pathname [2] for the build tree.
The initial CD safeguard has been applied to the `buildconf` and
four more scripts:
- build/genif.sh
- scripts/dev/credits
- scripts/dev/genfiles
- scripts/dev/makedist
Rationale:
Cd-ing into the project root should always prematurely exit w/ FAILURE
as a required precondition for its invocation has not been met. This
should never go unnoticed as it always requires user intervention.
Similar and more specifically to the PHP build on *nix systems, the
grep command is required early to obtain the `php_extra_version` from
configure.ac. Previously, if the grep command is missing (or failing
due to not matching the line with the AC_INIT macro [1]), the internal
dev parameter would always be zero (0) which can easily result in the
situation that the configure script is not being rebuilt. This is
cumbersome as the rebuild of a configure script is more likely required
with checked-out dev versions under change rather than an already
properly set-up build environment on a dedicated build or release
system. Missing the fact that either the grep utility is missing or
the expectation of having the AC_INIT macro in configure.ac is unmet
should never go unnoticed as it always requires user intervention.
[1]: https://www.gnu.org/software/autoconf/manual/autoconf-2.67/html_node/Initializing-configure.html
[2]: https://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_chap03.html#tag_03_271
Closes GH-16717.
When passing an array, the key entry can get converted to a string if it
is an object, but this actually modifies the original array entry.
The test originally outputted:
```
array(2) {
[0]=>
string(...) => ...
[1]=>
string(0) ""
}
```
This is unexpected. Use zval_try_get_string() to prevent this behaviour.
Closes GH-16693.
For rename_variation12.phpt this is actually not necessary, since there
is no rename_variation11.phpt, but we still fix it to be in sync with
rename_variation12-win32.phpt which actually is prone to parallel
conflicts.
(cherry picked from commit d4263ddc40)
It is assumed that the serialization always had initialised its buffer
zend_string, but in the case of a type not serialising, it is null.
close GH-16599
zlog_buf_prefix() can return a larger length than what actually was
written due to its use of snprintf(). The code in
zlog_stream_prefix_ex() does not take this into account, other callers
do. What ends up happening then is that stream->length is set to the
length as if snprintf() was able to write all bytes, causing
stream->length to become larger than stream->buf.size, causing a
segfault.
In case the buffer was too small we try with a larger buffer up to a
limit of zlog_limit. This makes sure that the stream length will remain
bounded by the buffer size.
This also adds assertions to make the programmer intent clear and catch
this more easily in debug builds.
Closes GH-16680.
Additionally fixes wrong behaviour in ReflectionParameter when you first
have a construction that uses an object and the subsequent doesn't.
Closes GH-16672.
`timelib_astro_rise_set_altitude()` is not prepared to deal with non-
finite values (`nan`, `inf` and `-inf`) for `lon` and `lat`; instead
these trigger undefined behavior. Thus we catch non-finite values
before even calling that timelib function; for `date_sun_info()` we
trigger `ValueError`s; for `date_sunrise()` and `date_sunset()` we
silently return `false`, since these functions will be sunsetted
anyway.
Closes GH-16497.
These have been introduced a while ago[1], but their initialization has
been overlooked. Since we cannot rely on TLS variables to be zeroed,
we catch up on this.
[1] <e3ef7bbbb8>
Co-authored-by: Ilija Tovilo <ilija.tovilo@me.com>
Closes GH-16658.
A previous bug fix[1] relied on ODBC drivers to properly count down the
`StrLen_or_IndPtr` argument for consecutive calls to `SQLGetData()`.
Apparently, not all drivers handle this correctly, so we cannot assert
they do. Instead we fall back to the old behavior for drivers which
would violate the assertion.
A test against SQLServer (which we currently use in CI) would not make
sense, since the respective drivers do not exhibit that behavior.
Instead we target the regression test especially to a MS Access
database.
Since there is apparently no way to easily create an MS Access database
programmatically, we commit a minimal empty DB which is used for the
regression test, and could also be used by other test cases.
[1] <bccca0b53aa60a62e2988c750fc73c02d109e642>
Closes GH-16587.
This PR removes most of the logic of nightly_matrix.php by converting
nightly.yml to a workflow_call, and invoking it multiple times for each branch
in a new root.yml workflow. The naming of the files is intentionally kept to
make the diff readable. They may still be renamed afterward.
Closes GH-16642
