Re: [Fwd: Re: [Evolution-hackers] Question about defunc processes produced by libsoup on Solaris]

[Dan Winship <danw ximian com>]

On Mon, 2003-03-31 at 10:39, Alex Jiang wrote:
> Hi
> 
> Could u give more detailed explaination of why so called "zombies" are
> meaningful for soap?
> 
> I've never viewed the source of libsoap or used libsoup before.
> 
> But why zombies may happen in sol we should call wait to release them,
> while not in linux?

It's not a solaris vs linux thing.

The patch you submitted makes soup try to clean up *all* zombies when an
ssl proxy exits. That's inappropriate behavior for a library. Some of
those zombies may have been created by the application, or by other
libraries, which may be planning to wait() on them later to find their
exit status.

The bug here is that soup_ssl_hup_waitpid is not being called for every
ssl connection for some reasonOn Mon, 2003-03-31 at 10:39, Alex Jiang
wrote:
> Hi
> 
> Could u give more detailed explaination of why so called "zombies" are
> meaningful for soap?
> 
> I've never viewed the source of libsoap or used libsoup before.
> 
> But why zombies may happen in sol we should call wait to release them,
> while not in linux?

It's not a solaris vs linux thing.

The patch you submitted makes soup try to clean up *all* zombies when an
ssl proxy exits. That's inappropriate behavior for a library. Some of
those zombies may have been created by the application, or by other
libraries, which may be planning to wait() on them later to find their
exit status.

The bug here is that soup_ssl_hup_waitpid is not being called for every
closed ssl connection for some reason, and that's what needs to be
fixed.

-- Dan