Disable resource-heavy tests by default (>1GB of memory usage), unless the
RUN_RESOURCE_HEAVY_TESTS env variable is set.
Fixes GH-20762
Closes GH-20935
When installing executables that were built using libtool, we are
supposed to use
$ libtool --mode-install <install-command>
rather than the bare <install-command>. This is discussed ever so
briefly in the "Installing executables" section of the GNU libtool
documentation:
https://www.gnu.org/software/libtool/manual/libtool.html
So far this has not caused a problem with GNU libtool on the platforms
that PHP supports, but there is an alternate libtool implementation
called slibtool that stores wrappers at the locations where PHP is
expecting the true executables to live. As a result, the wrappers (and
not the executables) are installed when slibtool is used to build PHP.
This is fixed by replacing,
$(INSTALL)
with
$(LIBTOOL) --mode=install $(INSTALL)
in the install-foo rules for the executables that are built with
libtool.
Closes GH-13674
The aim of this PR is twofold:
- Reduce the number of highly similar TMP|VAR handlers
- Avoid ZVAL_DEREF in most of these cases
This is achieved by guaranteeing that all zend_compile_expr() calls, as well as
all other compile calls with BP_VAR_{R,IS}, will result in a TMP variable. This
implies that the result will not contain an IS_INDIRECT or IS_REFERENCE value,
which was mostly already the case, with two exceptions:
- Calls to return-by-reference functions. Because return-by-reference functions
are quite rare, this is solved by delegating the DEREF to the RETURN_BY_REF
handler, which will examine the stack to check whether the caller expects a
VAR or TMP to understand whether the DEREF is needed. Internal functions will
also need to adjust by calling the zend_return_unwrap_ref() function.
- By-reference assignments, including both $a = &$b, as well as $a = [&$b]. When
the result of these expressions is used in a BP_VAR_R context, the reference
is unwrapped via a ZEND_QM_ASSIGN opcode beforehand. This is exceptionally
rare.
Closes GH-20628
The zend_exception_save() and zend_exception_restore() functions were removed in
GH-20256. However, the fuzzer expects to be able to call opcache_invalidate()
even if there was an exception thrown. I'm not sure why exactly, but let's
restore the previous behavior.
Closes OSS-Fuzz #476466130
These are leftovers from the pre-PHP-7.0 era. This also implicitly solves
GH-20564 by not clearing exceptions before entering the autoloader.
Closes GH-20256
Fixes GH-20564
The arg_info member of zend_function is now always a zend_arg_info*. Before,
it was a zend_internal_arg_info* on internal functions, unless the
ZEND_ACC_USER_ARG_INFO flag was set.
Closes GH-19022
According to @cmb69, PHP does not require VLA support
(https://github.com/php/php-src/pull/10304#discussion_r1069343092).
VLAs are a bad idea for several reasons, so let's get rid of them.
Two of the VLAs were probably unintended; unlike C++, C doesn't have
the concept of "constant expressions", so an array with a "const"
length is technically still a VLA. This is fixed by removing the
"const" variable, and using sizeof() instead.
(cherry picked from commit ff2a211d55)
This aligns the behaviour with what the stubs say.
And even if one fixes the stubs the behaviour is not identical due to missing indirect handling.
This indicates that using objects is never done, so do the easy fix of changing the ZPP specifier.
Closes GH-20465
The stubs say array so in debug mode we get a ZPP violation assertion and even by fixing the stubs the behaviour is not identical due to missing indirect handling.
This indicates using objects was never done, thus use the correct ZPP specifier
> LeakSanitizer has encountered a fatal error.
This happens only on 8.2/8.3 for some reason. Don't merge for 8.4 for now, we
can do that later if it reoccurs.
There have been misterious macOS failures with the following error for a long
time, which would usually happen 1-2x per nightly run:
> The hosted runner lost communication with the server. Anything in your
> workflow that terminates the runner process, starves it for CPU/Memory, or
> blocks its network access can cause this error.
After way too much debugging, it looks like this is caused by the two fpm tests
skipped in this commit. When there's a failure, the responsible test will run
for at least 120 seconds until the job is eventually killed by GitHub Actions.
It's not clear yet why the tests stall.
Debugging this is a PITA because the GitHub Actions log gets partially lost for
killed jobs. It took an external log server to actually figure out where the job
fails. Let's disable the tests first to be sure this actually solves the issue.
The fix fixes some other races that could result in mixed up copy and
nprocs number. It requires creating a copy in a similar way like for
request status.
This was not previously used to not impact other requests. However this
does not make that much sense as the only thing that impacts it is
holding the lock and not waiting for it. It is true that if there was a
big contention then the lock would be acquired more often but that can
be achieved by using fpm_get_status in loop so it should not make a
huge difference hopefully.
Closes GH-19974
This changes default for fastcgi.script_path_encoded INI to have
default behavior without a BC break. There has been already issue
about this in RC so it is very likely that it could have much bigger
impact so it's better to be safe.
