Inotify causing silent merge corruptions ?!

[Marc Bevand]

Marc Bevand <m.bevand <at> gmail.com> writes:
> 
> This occured with Mercurial at revision 11660d2918af (1.1.1 + some
> trivial changes), running on Linux kernel 2.6.27, with the inotify
> extension, on a large repo (700MB, 100k files) residing on a local XFS
> filesystem. I suspect an inotify bug, maybe a race condition. I cloned
> my repo to be in the state it was at right before the merge, to
> reperform the merge, but this time it does merge as expected. Next thing
> I am going to try is to write a script to repeatedly attempt the merge.
> If it is a race condition in inotify, it should eventually reproduce the
> bug.

I wrote that script and it does reproduce the merge corruption everytime on my 
repo! As I predicted it seems to be a race condition: if hg merge is 
immediately succeeded by hg commit (with no pause in between), some files will 
not be merged at all (as if inotify was still updating the dirstate or 
something in the background). If I either insert a sleep 5 between the two 
commands, or disable inotify altogether, everything is fine.

I am now trying to build a test case to allow anyone to reproduce the bug, but 
the race condition seems to be dependent on the complexity of the repo (in my 
case thousands of files were being merged). And so far my attempts to build 
test repos with thousands of dummy files failed to reproduce the bug.

-marc