Re: g_spawn question

[Havoc Pennington <hp redhat com>]

Thomas Vander Stichele <thomas urgent rug ac be> writes:
> I have a question regarding the g_spawn functions in glib.  I'm writing an 
> infra-red panel application using gnome2 libraries.
>
> One of the things this app does is execute external helper shell scripts.  
> These scripts output minimized information while they're running.  This 
> output should then be captured by the panel app and displayed line by 
> line.
>
> Now, what I did was use g_spawn_async_with_pipes and connect a GIOChannel 
> to the stdout of the spawned process.  This works; I can process the 
> output of the spawned process line by line.
>
> The problem starts when the spawned process is done running.  For some 
> reason my app starts spinning idly, using 99.9% CPU.
>
> Now, my suspicion is that when the spawned process stops, the 
> corresponding iochannel of stdout isn't unreffed.  I'm not sure if this is 
> the case or how to verify this; I can only tell from what I see and from 
> my (maybe mistaken) interpretation of the glib source code.  But I suppose 
> it might be likely that, for example, the file descriptor of the spawned 
> stdout is closed, but the iochannel doesn't know.
>
> So, my question is, does this make sense ? Is this the right behaviour ? 
> Is there some way I can install a callback to be called when the spawned 
> process stops ? How can I cleanly prevent my app from spinning idly, or 
> otherwise debug the situation ?

Yes, I think you diagnosed this correctly. If you haven't figured it
out in the interval since you posted this mail (moderation queue was
lagging I guess), you need to remove your GIOChannel when the other
end of the pipe is closed. I think you get a G_IO_HUP condition when
this happens, IIRC.

Another option is to use G_SPAWN_DO_NOT_REAP_CHILD or whatever it's
called and then you get SIGCHLD when the child exits.

Havoc