Races, corruption and segfaults; was: Stress-Test the Mercurial revlog; was: Hash collision

[Arne Babenhauserheide]

Hi, 

I now ran the test together with Bill Barry, and we now have almost 8000 
revisions in the repo, in about 3MiB space. 

We added something between a commit a second and a commit every 2 seconds. 

While doing to, we experienced pretty bad races (Bill not being able to get 
changesets in for a long time, because I just had the far lower ping time (to 
my own computer, and he from the USA to Germany), but these were expected and 
just led to some lost time and automatic merges. 

We saw some other strange artifacts, though: The repository got corrupted a 
few times - I moved the damaged versions away. Most were with the scm version 
of Mercurial, but the last one was with mercurial 1.2.1. 

I then did a test with local pushing instead of using the wire protocol, and 
there I got segfaults in the pusher, but no corruption (the locks worked). 

Also we saw the robustness of the DVCS approach: When the main repository got 
corrupted, we couldn't push anymore, but no history was lost, and once I had 
initiated the push repo again, the test just ran on. 

Should I open a bug and attach a corrupted repository? 

Best wishes, 
Arne

--- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- 
   - singing a part of the history of free software -
              http://infinite-hands.draketo.de
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part.
Url : http://selenic.com/pipermail/mercurial/attachments/20090430/f1fb156b/attachment.pgp