[darcs-devel] [issue1927] It's too late to use atexit

Eric Kow bugs at darcs.net
Wed Aug 25 17:32:38 UTC 2010


Eric Kow <kowey at darcs.net> added the comment:

On Fri, Aug 20, 2010 at 08:23:14 +0000, Simon Marlow wrote:
> I just did a 'darcs get' of Ganesh's rebase branch, and followed the 
> suggestion to hit ^C to get a lazy repository.  Here's what happened:

This looks like it might be a duplicate of
http://bugs.darcs.net/issue1914 which Simon PJ reported
but it's slightly hard to tell.

Also, I can't reproduce this with darcs HEAD

> It's too late to use atexit

The fact that this happens 18 times and not 1 should be a nice clue.
What does Darcs do frequently enough on atexit time that be what we're
looking for?

> /home/simonmar/.darcs/cache/patches/0000032950-601def3f87db9fdd15d4da9ec4e5e973e0ad12c0cc7bec9b89842ae6ed5b3d4e-new_6ec539135c824a59b72c5a3f463c4d66: 
> getSymbolicLinkStatus: does not exist (No such file or directory)

This is also yucky, and I also can't reproduce it (but I dimly recall
seeing things like this before and hearing they are essentially
harmless).

Note that issue1804 is slightly related to this.  It won't solve the
underlying problem, but it could make it slightly less frightening.

BTS training for future darcs Issue Managers:

1. I've marked this bug waiting-for because I think there is
   no more way to progress on it without some specific 'outside'
   constraint being fulfilled.  In this case, the constraint is
   progress on issue1914.
2. I was also tempted to split this into two tickets, but then what's
   holding me back is the possibility that it's actually just the same
   underlying bug.  So we'll have to do more BTS and debugging work
   before we do the split.

-- 
Eric Kow <http://www.nltg.brighton.ac.uk/home/Eric.Kow>
For a faster response, try +44 (0)1273 64 2905 or
xmpp:kowey at jabber.fr (Jabber or Google Talk only)

__________________________________
Darcs bug tracker <bugs at darcs.net>
<http://bugs.darcs.net/issue1927>
__________________________________


More information about the darcs-devel mailing list