Edit report at https://bugs.php.net/bug.php?id=44942&edit=1
ID: 44942
Updated by: pajoye@php.net
Reported by: inqualab1985 at gmail dot com
Summary: exec() hangs apache
Status: Assigned
Type: Bug
Package: Program Execution
Operating System: Windows 2000 SP4
PHP Version: 5.2.5
Assigned To: pajoye
Block user comment: N
Private report: N
New Comment:
This should be fixed already. Please try latest 5.6
Previous Comments:
------------------------------------------------------------------------
[2015-03-23 21:18:13] yardsticks at gmail dot com
Same issue here, years later... I built a whole script around the expectation that this would work.
The solution posted did not work for me either :(
------------------------------------------------------------------------
[2012-12-06 12:54:57] pajoye@php.net
See also #44994
------------------------------------------------------------------------
[2012-12-05 12:04:02] claudix dot kernel at gmail dot com
Seen the same behavior, not only in exec(), but also in similar functions as proc_open/proc_close.
When there are concurrent scripts during a same PHP session, the script spawning the process
randomly hangs.
System:
- Windows 2003 server SP2
- Apache 2.2.22/ PHP 5.4.3
I can confirm that calling session_write_close() and then session_start() does the trick.
I've observed, though, that the code below doesn't work:
$proc = proc_open($cmd,$pipedesc,$pipes);
//do stuff with pipes...
//... and close pipes
session_write_close(); //Close session before hanging function
$retval = proc_close($proc);
session_start(); //restore session
But the code below *does* work:
session_write_close(); //Close the session before proc_open()
$proc = proc_open($cmd,$pipedesc,$pipes);
//do stuff with pipes...
//... and close pipes
$retval = proc_close($proc);
session_start(); //restore session
This made me go into the PHP source code (actually the source file "proc_open.c").
I've noticed that the command passed to proc_open() is spawned by calling the WINAPI function
CreateProcess(...) with the parameter "bInheritHandles" set to TRUE. As of MSDN
documentation, if this parameter is TRUE then all handles are inherited by the child process. It
seems that the handle of the session is being inherited by the child process but for some reason the
OS doesn't release it when the process ends, eventually yielding a deadlock. The code snippets
above show this: the session has to be closed before calling proc_open() to prevent the spawned
command from inheriting the session handle. People using exec() cannot see this effect because
exec() virtually embeds proc_open/proc_close.
May this give a clue to PHP developers?
Claudi
------------------------------------------------------------------------
[2012-10-15 15:00:52] mail at GerhardBechtold dot com
Pajoye,
I didn't find any documentation on the service sensitivity of the new PHP.
You might be right, that the eof of the console stream is not taken care properly, but in my case
the system is now working (again).
------------------------------------------------------------------------
[2012-10-15 09:12:30] pajoye@php.net
@mail at GerhardBechtold dot com
this is documented, the shell/exec permissions have to be given.
However I do not think it is related to the original issue which is caused by a
real bug in the php stream, where the eof of the console stream is not correctly
detected and ends in an endless loop.
------------------------------------------------------------------------
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=44942
--
Edit this bug report at https://bugs.php.net/bug.php?id=44942&edit=1