Re: RFS: adun.app (updated package)

[Yavor Doganov <yavor@gnu.org>]

Andreas Tille wrote:
> 
> It is *really* difficult?

The difficult part is figuring out what to clean and when.  Many
processes can write to $TMP/GNUstepSecure$UID.

> My first idea was to write a wrapper that checks any remainings in
> /tmp and just remove what not should be there.

The thing is, it is not possible to establish in a reliable way what
should and what should not be there.  The GNUstep Base library itself
doesn't know this, and it's documented.  There are helper functions
that clean stuff (e.g. remove ports and the connection map table), but
they rely on the fact that everything (tool, app, etc.) shuts down
cleanly, which has never been the case.

Nevertheless, it is worth trying.  I already patched gnustep-base on
all my machines and I'll see if I can reproduce the bug.

> I admit that this is a stupid brute force 

If it is the sole reason that is causing the problem, this approach is
OK in my book.

If someone reports a bug for adun.app that resembles this issue, I
kindly ask you not to advise him to remove the temporary files.  I'd
like to ask this hypothetical person to try my patch first (I'm
subscribed to the PTS, FYI).

Thanks once again for everything, it's been a great pleasure having
this conversation with you.