oops again...
cstb wrote:
>
>
> I'm a little rusty here, but this sounds like a fork?
>
VW doesn't use fork...clone then execve.
>
> Ok, so it must be a fork.
>
See above.
>
> Hmmm. I'd expect parent
> (which is waiting
> for children
> -- ALL children --
> to exit
> )
> would continue waiting until grandchild completes.
> Unless/until the grandchild is emancipated,
> which is *not* the usual thing.
>
This isn't my understanding of it at all. Grandchild is orphaned when
Child exits. It becomes a child of init, which will "wait" on it when
it exits. "ps exjf" verifies this by showing process 1 as the "parent
pid" of the grandchild.
> A simple fix would be to <bash> your example
> into working the way you intend:
>
> Parent spawns child via #cshOne:.
> Child spawns grandchild via something like
>
> cshOne: 'bash -c ''originalGrandchild &'''
>
>
OK, I tested this (but I used shOne:, not cshOne: as I don't have csh on
my box). The behavior in this case is the same as my original. Parent
sits waiting.
> Or, you could look up process groups,
> and methods for escaping therefrom,
> and find a way to convey emancipation
> after a fork, from within VW.
>
>
> CAVEAT - I don't know if your OS variant
> actually does anything about process groups.
>
>
Yes, Linux support process groups and sessions. If I look at the
groupid and sessionid of the grandchild (in my sample code) I see that
they indeed match its PID so it is in its own group/session. In other
parts of the thread you will see discussion that Linux delivers SIGCHLD
for Child to Parent which VW handles properly. The problem seems to be I/O.
David
_______________________________________________
vwnc mailing list
[hidden email]
http://lists.cs.uiuc.edu/mailman/listinfo/vwnc
Locked
