Re: git pull aborts in 50% of cases

H. Peter Anvin <hpa <at> zytor.com> writes:

Actually, it turns out the two servers were running different versions; 
one 0.99.9j and one 0.99.9k.  They're both running 0.99.9j now.

0.99.9k is clearly bad.

This is troublesome, since I do not think it is hitting the one that was
causing trouble for the snapshotting procedure, and I cannot seem to reproduce
it with random set of flags and parameters myself.

I know you run it with --inetd, but what other parameters do you run it with,
and in what kind of environment?  Specifically:

 - do you use whitelist, like "git-daemon --inetd --export-all /pub/scm"?
 - if so, are any symlinks involved in /pub area?
 - if so, does adding real path like /mnt/real/pub/scm for /pub/scm help?
 - do you run with --strict-paths?