Description of problem: "man 3 exit" says, amongst other: ... The exit() function causes normal process termination and the value of status & 0377 is returned to the parent (see wait(2)). All functions registered with atexit(3) and on_exit(3) are called, in the reverse order of their registration. ... However, the atexit() hooks are never called if you invoke any of the exec*() functions. Version-Release number of selected component (if applicable): glibc-2.8-8.x86_64 How reproducible: Always Steps to Reproduce: 1. Copy the example from 'man 3 atexit' to test.c 2. Modify it so that instead of "exit(EXIT_SUCCESS)" at the end we have, for example: execl ("/bin/ls", "ls"); 3. Compile and run the example Actual results: $ ./a.out ATEXIT_MAX = 2147483647 a.out test.c Expected results: $ ./a.out ATEXIT_MAX = 2147483647 That was all, folks a.out test.c Additional info: I think it is a serious bug as it does not allow applications that rely on it to cleanup things that won't automatically cleanup during a normal process destruction.
I tried on a micro linux-from-scratch variant I have running here in a virtual machine, and the bug is still there, so it's really a upstream bug. I have copied the report to the bugzilla recommended on glibc homepage, and added a link to that bugzilla.
I was told by Ulrich Drepper that this is behaviour is conformant to standards. I found in the "The Open Group Base Specifications issue 6" the page related to exec() here: http://www.opengroup.org/onlinepubs/000095399/functions/exec.html The only quote mentioning atexit is: <<After a successful call to any of the exec functions, any functions previously registered by atexit() or pthread_atfork() are no longer registered.>> This is somehow vague, but anyway I think this can be re-classified as a bug in the man page. FreeBSD and Solaris man pages just repeat the same quote from standard. Changing target package to man-pages-2.78-2.fc9.noarch.
The description details from which results the fact - the atexit stuff is not called before the execl call starts - are on execve page (there is link from execl page): "All process attributes are preserved during an execve(), except the following: .. * Exit handlers are not preserved (atexit(3), on_exit(3)). .. " It seems for me to be sufficient info so I think there is no need to add another notice about it.
Indeed, there's a note on execve() page. I do not insist, but I think it would be helpful to note this on exec page as well. And even emphasize that the handlers are not called, not just "not preserved". Maybe I don't understand it well, but "handlers are not preserved" and "handlers are not called and not preserved" sound a bit different to me. Anyway, this bug can be closed I think.