Mercurial bug? Core hg dev input requested??? (was Re: win32text extension processing 00changelog.i?)

[Benoit Boissinot]

On Tue, Oct 13, 2009 at 11:55 AM, Ringo De Smet <ringo.desmet at gmail.com> wrote:
> More information...
>
> 2009/10/13 Ringo De Smet <ringo.desmet at gmail.com>:
>>
>> I am tempted to say that there is a bug of some sort in the Mercurial
>> internals, but it's very hard to track, as the problem is still not
>> consistently reproducible. Input from a core Mercurial developer would
>> be of great help!
>
> Attached are two parts of the server side traceback. The first file
> (win32text.txt) shows the error I got when the win32text extension was
> active, the other file is the error I get now with the merge_direction
> extension active. This is an extension posted recently by another
> Mercurial user that I adapted for our branching scenario.

Ok, if it's possible can you disable those hooks?
Did you check "hg verify" (client and server)?
Can you send the 00changelog.{i,d} files after the push with hooks disabled?

(what is your hg version? I would be nice to have the resulting
changelog files from a failed push, usually a new file ending with .a
is created, could you try to save it, for example you could have the
time to do it when running a hook that just sleeps for many seconds)

regards,

Benoit