[tac_plus] Re: opt_G broken again :(

[john heasley]

Tue, Mar 17, 2009 at 07:27:49AM -0400, John Payne:
>
>
> On Mar 17, 2009, at 3:36 AM, john heasley <heas at shrubbery.net> wrote:
>
>> Mon, Mar 16, 2009 at 02:06:40PM -0400, John Payne:
>>>
>>> On Mar 16, 2009, at 9:53 AM, John Payne wrote:
>>>
>>>> The opt_G timeline:
>>>>
>>>> First patch: http://www.shrubbery.net/pipermail/tac_plus/2007-October/000158.html
>>>> Second patch to fix defunct processes: http://www.shrubbery.net/pipermail/tac_plus/2007-October/000163.html
>>>> Third patch to make it work: http://www.shrubbery.net/pipermail/tac_plus/2008-June/000246.html
>>>>
>>>> Unfortunately, the third patch has started causing defunct processes
>>>> again.
>>>
>>> This seems to resolve the defunct process issue:
>>>
>>> *** tac_plus.c.orig    2009-03-16 14:05:29.000000000 -0400
>>> --- tac_plus.c    2009-03-16 14:05:49.000000000 -0400
>>> ***************
>>> *** 447,452 ****
>>> --- 447,457 ----
>>>            */
>>>           open_logfile();
>>>       }
>>> + #ifndef REAPCHILD
>>> +         signal(SIGCHLD, reapchild);
>>> + #else
>>> +         signal(SIGCHLD, SIG_IGN);
>>> + #endif /* REAPCHILD */
>>>       }
>>>
>>>       ostream = NULL;
>>>
>>>
>>> Probably a cleaner way, but... it works for me :)
>>
>> What O/S is this?
>>
>
> Linux 

Grumble.  The default position of SIGCHLD should be SIG_IGN.