[darcs-users] Darcs segfault

Petr Rockai me at mornfall.net
Tue May 17 13:44:44 UTC 2011

Hi Erik,

Erik de Castro Lopo <mle+tools at mega-nerd.com> writes:
> Ganesh Sittampalam wrote:
>> >> Is this a known issue?
>> >
>> > Sorry, darcs 2.4.4 from Debian.
>> Almost certainly caused by _darcs/index - try just removing that.

>> I'm not sure if later versions make a check on endianness/bit size before 
>> reading the file, but we should add one if not.
> Yes, that is a bug that I would love to see fixed.

It would be great if you could do the same kind of test with darcs 2.5
(i.e. copy a darcs-2.5-created index over to a BE platform). The index
is supposed to be byte-order neutral (it should use LE everywhere on all
platforms). But bugs may have crept in, since I don't own a BE machine
to test. (Bugs almost certainly meaning a missing xlate somewhere.)

(Hashed-storage 0.4.9 has seen a BE-related fix, but Debian has 0.4.13,
so presumably darcs is linked against that. We are likely facing a
different bug that is still unfixed :|.)


PS: If you could take an identical (including mtimes) darcs repository
on both BE and LE platform and share the resulting _darcs/index files
from both, that should help me fix the problem. (Both rsync and NFS
should be able to give you the required repository clone.)

id' Ash = Ash; id' Dust = Dust; id' _ = undefined

More information about the darcs-users mailing list