Edit report at https://bugs.php.net/bug.php?id=60978&edit=1
ID: 60978
Updated by: thekid@php.net
Reported by: thekid@php.net
Summary: exit code incorrect
Status: Open
Type: Bug
Package: Scripting Engine problem
Operating System: Windows
PHP Version: 5.4.0RC7
Block user comment: N
Private report: N
New Comment:
...oh, and:
* Uncaught exception exit, php -r 'throw new Exception("test");' ; echo
$? = 255
Previous Comments:
------------------------------------------------------------------------
[2012-02-05 09:59:54] thekid@php.net
Your patch breaks the exitcode 254 for parse errors.
We have the following cases and expected exit codes:
* Clean exit, php -r 'echo "";' ; echo $? = 0
* Clean explicit exit with message, php -r 'exit("test");' ; echo
$? = 0
* Clean explicit exit with exitcode, php -r 'exit(2);' ; echo $? = 2
* Exit after a fatal error, php -r '$fatal->error();' ; echo $? = 255
* Exit after compile error, php -r '$->' ; echo $? = 254
------------------------------------------------------------------------
[2012-02-04 20:43:08] php-dev at zerocue dot com
Alright, in 5.3.10@php_cli:1023 the call to zend_eval_string_ex() never returned
after a zend_bailout(), if it had this bug would have been present there because
zend_bailout always returns FAILURE even in non-failure cases.
In 5.4 and trunk, the zend_eval_string_ex() does return and hits the
exit_status=254 line. The proper fix would be to have zend_bailout not return
FAILURE unless it did fail but this is a much larger change. In this case the
patch simply ignores the return state of the zend_eval_string_ex() and always
returns the EG(exit_status) value.
------------------------------------------------------------------------
[2012-02-04 20:00:43] php-dev at zerocue dot com
This does not occur when executing a script file, only through -r.
------------------------------------------------------------------------
[2012-02-04 19:59:37] php-dev at zerocue dot com
Actually my last post has nothing to do with it, mis-read the code. I don't see
where the problem is, the executor_globals.exit_status is being set to two
properly and the LONGJMP(bailout) still has the right exit_status value in the
executor.
I did confirm that this was fine in 5.3.10 but not in 5.4 or trunk.
------------------------------------------------------------------------
[2012-02-04 19:44:24] php-dev at zerocue dot com
Looks like this (on my system) is related to ZEND_VM_SPEC not being defined.
zend_vm_def.h: 4567
#if !defined(ZEND_VM_SPEC) || (OP1_TYPE != IS_UNUSED)
Results in all variants of ZEND_EXIT to become: #if 0 || ...
There are only two references to ZEND_VM_SPEC in the code base, neither are
defines. What does this control, where should it be defined?
------------------------------------------------------------------------
The remainder of the comments for this report are too long. To view
the rest of the comments, please view the bug report online at
https://bugs.php.net/bug.php?id=60978--
Edit this bug report at https://bugs.php.net/bug.php?id=60978&edit=1