For master (8.4-dev) I merged GH-13381. But that PR changes public API
of TSRM, so cannot be used on lower branches.
This patch is a safe workaround for the issue, in combination with a
pre-existing fix using `ifdef ZTS + if (module_started)` inside pgsql
and odbc. The idea is to delay unloading modules until the persistent
resources are destroyed. This will keep the destructor code accessible
in memory.
This is not a proper fix on its own, because we still need the
workaround of not accessing globals after module destruction.
The proper fix is in master.
Closes GH-13388.
Since PHP 8.3, object handlers may be changed by setting
ce->default_object_handlers, rather than in ce->create_object. Some checks need
to be extended to check for the default handlers.
Closes GH-13272
* Fix GH-13193: Significant performance degradation in 'foreach' starting from PHP 8.2.13 (caused by garbage collection)
* Don't run zend_gc_remove_root_tmpvars() if GC is not active or GC buffer is empty
When investigating another bug reported by GitHub user 'tstangner',
I discovered that PHP segfaults when the INI parameter
zend.script_encoding is set to "pass". This bug dates back to
December 2022 (caused by yours truly in 953864661a).
If any PHP users in the wild were actually setting zend.script_encoding
to "pass" (which would be an utterly useless thing to do), I expect that
someone would have filed a bug report by now. The absence of such bug
reports is evidence that nobody is doing this.
Hence, it seems that the best fix is simply to disallow "pass" as a
choice for zend.script_encoding. The internal function
'php_mb_zend_encoding_list_parser' which I am modifying to accomplish
this has no other in-tree callers, aside from the 'exif' extension.
Further, exif only calls the function with a few hard-coded values, and
none of them are the string "pass", so this change will not have any
impact on exif.
zend_compile has an exception to this rule for constructors using
`zend_is_constructor`, which compares the function name to
`__construct`. Sadly, `zend_is_constructor` is not a public API, but we
can just do the string compare ourselves.
Closes GH-13179.
* Arrays returned from range are never empty
* When step is a double value representable by a long, it is coerced implicitly.
As such, passing a double step no longer guarantees that the result is a
non-int array.
Closes GH-13166
Recentlty this insructions were updated to emit warning on inability to
convert double index to long. This may lead to exception.
This fixes memory leak on wordpress test suite (nightly workflow)
ts_resource() and php_request_startup() both eventually call zend_max_execution_timer_init(), which didn't have a guard to prevent recreating timers, thus resulting in leaking timers. This adds a guard to prevent the leak.
This is a false positive. The cycle detection code stops at 10.000
iterations. Instead of stopping at a fixed amount, make it more robust
by implementing Floyd's cycle detection algorithm.
Closes GH-12954.
This regressed in cd53ce838a.
The loop with `zend_hash_iterators_update` hangs forever because
`iter_pos` can't advance to idx. This is because the
`zend_hash_iterators_lower_pos` upper bound is `target->nNumUsed`,
but that is set to `source->nNumOfElements`.
That means that if there are holes in the array, we still loop over all
the buckets but the number of bucket slots will not match.
Fix it by changing the assignment.
Closes GH-12831.
This has been the case at least since PHP 5.4. Thanks to Girgias for
pointing it out.
It appears that there are several global variables internal to mbstring
which can be queried via mb_get_info() and which could be NULL, but
at the very least, we know that "mbstring.http_input" is one of them.
