[Pidgin] #1496: Signal handling crashes some non-Linux platforms

[Pidgin]

#1496: Signal handling crashes some non-Linux platforms
--------------------------+-------------------------------------------------
 Reporter:  psfales       |       Type:  defect
   Status:  new           |   Priority:  minor 
Component:  pidgin (gtk)  |    Version:  2.0.1 
 Keywords:                |    Pending:  0     
--------------------------+-------------------------------------------------
 I'm workin on porting pidgin to several non Linux platforms.   On several
 of them, the handling of SIGCHLD results in unlimited stack growth until
 the stack limit is reached and the program crashes.

 pigdin/gtkmain.c has code in sighandler() which looks like this:

         case SIGCHLD:
                 /* Restore signal catching */
                 signal(SIGCHLD, sighandler);
                 alarm(1);
                 break;
         case SIGALRM:
                 clean_pid();
                 break;

 The comments indicate that we want to wait one second before calling
 waitpid() to reap the child processes.  However, on some platforms
 (Solaris 2.5.1, HP/UX 10.20 and Irix 6.2) as soon as signal() is called,
 sighandler() is immediately called again to inform the process that not
 all of the children have been wait'ed on yet.  This results in
 sighandler() being continuosly called and  continuosly growing its stack.

 As a workaround, I tried moving the "signal(SIGCHLD, sighandler);" down
 inside the clean_pid() function after all the waitpid() calls have been
 completed (i.e the same place SIGALRM is re-enabled).  It appears to work,
 but I don't know the code well enough to say that all scenarios are
 covered.

 I've only actually verified this in 2.0.0, but the code looks the same in
 2.0.1

-- 
Ticket URL: <http://developer.pidgin.im/ticket/1496>
Pidgin <http://pidgin.im>
Pidgin