TLS is already used in TSRM, the way exporting the tsrm cache through
a thread local variable is not portable. Additionally, the current
patch suffers from bugs which are hard to find, but prevent it to
be worky with apache. What is done here is mainly uses the idea
from the RFC patch, but
- __thread variable is removed
- offset math and declarations are removed
- extra macros and definitions are removed
What is done merely is
- use an inline function to access the tsrm cache. The function uses
the portable tsrm_tls_get macro which is cheap
- all the TSRM_* macros are set to placebo. Thus this opens the way
remove them later
Except that, the logic is old. TSRMLS_FETCH will have to be done once
per thread, then tsrm_get_ls_cache() can be used. Things seeming to be
worky are cli, cli server and apache. I also tried to enable bz2
shared and it has worked out of the box. The change is yet minimal
diffing to the current master bus is a worky start, IMHO. Though will
have to recheck the other previously done SAPIs - embed and cgi.
The offsets can be added to the tsrm_resource_type struct, then
it'll not be needed to declare them in the userspace. Even the
"done" member type can be changed to int16 or smaller, then adding
the offset as int16 will not change the struct size. As well on the
todo might be removing the hashed storage, thread_id != thread_id and
linked list logic in favour of the explicit TLS operations.
* origin/master: (31 commits)
Fixed C++ incompatibility
update the certificate used for the test, as it expired recently
Fixed immutable arrays support
Fix counting of "R:" references in serialize()
Remove dead code
Test use($this) being an error
Move list() condition into assign_znode
typo
NEWS
Fix bug #68074 Allow to use system cipher list instead of hardcoded value
Avoid double checks
the order of the blocks should be Core, then exts in alphabetical order
add missing NEWS entry for the phpdbg compilation fix
add NEWS entry for #68088
Make QM_ASSIGN, JMP_SET and CAST return IS_TMP_VAR.
Removed useless helper
Drop unused result argument
Fix ct binding for cuf/cufa functions
Fix detection of write to built-in function for references
Test use of string names with \ prefix
...
- windows only, on linux gcc would cause a warning that the function returns,
as no noreturn functions are used
- ZEND_NORETURN works on windows as well now, using __declspec(noreturn),
which is useful for the other cases
- one more function call will be needed, which is not critical for the
error reporting
- the way is open to enable the same for gcc as soon as it's not causing a warning
* PHP-5.6:
update NEWS
Only destruct if EG(active) in zend_shutdown(). (bug #65463, #66036)
Fix typo from commit 32314f6b6
Fix destruction order in zend_shutdown (bug #65463, #66036)
Conflicts:
Zend/zend_compile.h
Zend/zend_execute_API.c
* PHP-5.5:
update NEWS
Only destruct if EG(active) in zend_shutdown(). (bug #65463, #66036)
Fix typo from commit 32314f6b6
Fix destruction order in zend_shutdown (bug #65463, #66036)
* PHP-5.4:
update NEWS
Only destruct if EG(active) in zend_shutdown(). (bug #65463, #66036)
Fix typo from commit 32314f6b6
Fix destruction order in zend_shutdown (bug #65463, #66036)
* pull-request/770:
Only destruct if EG(active) in zend_shutdown(). (bug #65463, #66036)
Fix typo from commit 32314f6b6
Fix destruction order in zend_shutdown (bug #65463, #66036)
Now one common stack to handle both, which stores znodes instead of
full oplines (foreach copy stack) or switch entries (switch cond
stack).
Also removed EG(start_op) while at it.
As far as I can discern these are leftovers of the interactive
shell implementation that was used before PHP 5.4. Now the readline
ext makes use of normal eval calls for this.
So, dropping these until there is evidence to the contrary, as they
currently wouldn't work anyway.
If Apache or a similar SAPI receives a signal during PHP processing
it calls zend_shutdown() without calling shutdown_executor().
#65463: If a module like Gearman or Memcached is loaded,
in the unfixed version it is unloaded by zend_destroy_modules() before the
CG(CLASS_TABLE) is destructed. When CG(CLASS_TABLE) is destructed,
any pointers to methods (specifically around destruction) in the unloaded
module's .so are now dangling and the process segfaults.
#66036: Any subclasses of an internal class like ArrayObject need
to be destructed in order: subclass first and then the internal class. In the
unfixed version zend_shutdown() clears the CG(CLASS_TABLE) from the head
of the list onwards, so internal classes are destructed first and user-defined
classes last. Internal classes are alloc/deallocated with malloc/free while
user-defined classes with emalloc/efree. If there's shared data between them
then efree() could be called instead of free() leading to a seg-fault.
