Edit report at https://bugs.php.net/bug.php?id=61873&edit=1
ID: 61873 User updated by: dcb at insomniacsoft dot com Reported by: dcb at insomniacsoft dot com Summary: proc_get_status always returns exitcode -1 Status: Open Type: Bug Package: Program Execution Operating System: Linux PHP Version: 5.4.1 Block user comment: N Private report: N New Comment: Ok, I think I've solved it. After some other tests I figured out it was a build config issue and in the end I narrowed it down to --enable-sigchild. I don't know why and without really understanding what it does I added that config option when I moved to php 5.4. It also breaks php 5.3. I've tested with php 5.3.6, 5.3.11 and 5.4.1 and after a clean build without --enable-sigchild proc_get_status will return the correct exitcode. I did a bit of research and found this article: https://blogs.oracle.com/opal/entry/php_oci8_signal_handling_and_e_1 Unfortunately I couldn't find the other bugs reporting this issue, but if they exist then this should be closed. Previous Comments: ------------------------------------------------------------------------ [2012-05-01 16:18:24] phpmpan at mpan dot pl Have you tried testing this on non-Ubuntu box? Maybe it's something that depends on Ubuntu configuration? I still don't like the fact that a second process is spawned after the one PHP is starting. The following code should also show calls to `wait` and values that are returned from processes. ------------------------------------------ strace -f php -r '$descriptorspec = array( 0 => array("pipe", "r"), 1 => array("pipe", "w"), 2 => array("file", "/tmp/error-output.txt", "a") ); $p = proc_open("/tmp/te", $descriptorspec, $pipes, "/tmp"); sleep(1); fclose($pipes[0]);fclose($pipes[1]); $s = proc_get_status($p); print_r($s);' 2>&1 | grep '/tmp/te\|wait' ------------------------------------------ ------------------------------------------------------------------------ [2012-05-01 01:12:19] dcb at insomniacsoft dot com Well, that is returning the proper exit code. It's strange that nobody else can reproduce this while I have it happening on 2 different Ubuntu machines. One is a 10.04 and the other one is 11.10. I was thinking that it might be something else, but keep in mind that on the same machine php 5.3 does work properly and returns the correct exit code in all the tests. ------------------------------------------------------------------------ [2012-05-01 00:28:56] phpmpan at mpan dot pl "and you the code is executing it. In other" should be: "and the code is executing it. In the other" Sorry ------------------------------------------------------------------------ [2012-05-01 00:27:25] phpmpan at mpan dot pl Since no one was able to reproduce this to this point, all that can be done is debugging it somehow on your side and eliminate possible reasons. What strace output tells us here is that "/tmp/te" actually exists and you the code is executing it. In other case the second call to `execve` would not return 0. If it is returning 0, it means that the interpretation of "/tmp/te" has actually started. He's my output for comparison (snaps for 5.4 and trunk): ------------------------------- [pid 7359] execve("/bin/sh", ["sh", "-c", "/tmp/te"], [/* 47 vars */] <unfinished ...> [pid 7359] execve("/tmp/te", ["/tmp/te"], [/* 46 vars */]) = 0 [pid 7359] open("/tmp/te", O_RDONLY) = 3 write(1, "/tmp/te", 7/tmp/te) = 7 ------------------------------- What is puzzling me is why your `sh` (pid:26199) is not replacing itself with the script. As we can see, another process (pid:26200) is spawned and it is the one that executes your script. My blind guess is that PHP receives status code from 26199, which - for some reason - is not waiting for 26200 or not passing the status code from it properly. Can you check what is this producing? ------------------------------- sh -c /tmp/te echo $? ------------------------------- Maybe the bug is not in PHP... ------------------------------------------------------------------------ [2012-04-30 23:24:42] dcb at insomniacsoft dot com Still getting -1 for /bin/ls as well I did the strace, here is output, unfortunately I have no idea how to interpret it: [pid 26199] execve("/bin/sh", ["sh", "-c", "/tmp/te"], [/* 19 vars */] <unfinished ...> [pid 26200] execve("/tmp/te", ["/tmp/te"], [/* 19 vars */]) = 0 [pid 26200] open("/tmp/te", O_RDONLY) = 3 write(1, "/tmp/te", 7/tmp/te) = 7 ------------------------------------------------------------------------ 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=61873 -- Edit this bug report at https://bugs.php.net/bug.php?id=61873&edit=1
