Re: signal handling problem

[christos%astron.com@localhost (Christos Zoulas)]

In article <201007111734.53291.mark%ecs.vuw.ac.nz@localhost>,
Mark Davies  <mark%ecs.vuw.ac.nz@localhost> wrote:
>On Sunday 11 July 2010 12:37:49 Christos Zoulas wrote:
>> >As written it fails to reap children (on 5.1_RC3) but if you remove the
>> >"if (sigchild)" prior to the "while ((pid = waitpid..." it works.
>> 
>> Looks like a kernel bug to me. Have you ktrace'd -i it?
>
>Well not "-i" but I have ktrace'd the parent:
>
>   460      1 cnid_metad CALL  pselect(1,0xbfbfe8f4,0,0,0,0xbfbfee38)
>   460      1 cnid_metad RET   pselect -1 errno 4 Interrupted system call
>   460      1 cnid_metad CALL  wait4(0xffffffff,0xbfbfee6c,1,0)
>   460      1 cnid_metad RET   wait4 6437/0x1925
>   460      1 cnid_metad CALL  close(2)
>   460      1 cnid_metad RET   close 0
>   460      1 cnid_metad CALL  wait4(0xffffffff,0xbfbfee6c,1,0)
>   460      1 cnid_metad RET   wait4 0
>   460      1 cnid_metad CALL  pselect(1,0xbfbfe8f4,0,0,0,0xbfbfee38)
>
>is what it looks like in the working (without if (sigchild)) version.  In 
>the original version (with if (sigchild)) its just
>
>   460      1 cnid_metad CALL  pselect(1,0xbfbfe8f4,0,0,0,0xbfbfee38)
>   460      1 cnid_metad RET   pselect -1 errno 4 Interrupted system call
>   460      1 cnid_metad CALL  pselect(1,0xbfbfe8f4,0,0,0,0xbfbfee38)
>
>so the pselect is getting interrupted its just that sigchild is not getting 
>set

Yes, that is why I am asking. It seems that while pselect is getting
interrupted by presumably the sigchld signal, the sigchld signal is
not being posted to the process so the sigchld variable is not set.

christos