Re: [egit-dev] EGit Corruption @ GitHub

["Shawn O. Pearce" <spearce@xxxxxxxxxxx>]

Scott Chacon <schacon@xxxxxxxxx> wrote:
> On Wed, Feb 3, 2010 at 3:26 AM, Shawn O. Pearce <spearce@xxxxxxxxxxx> wrote:
> > Hmmph. ??I wish GitHub helped us more with troubleshooting the
> > software bugs their customers are running into, rather than just
> > blaming us.
> 
> To be fair, it was an actual problem, it was specifically with EGit
> and I have (unsuccessfully) spent time trying to figure this out.  The
> problem is that most of our problems were with private repos that we
> can't use for research and many people using EGit are on Windows and
> generally not hugely sophisticated Git users that could help us debug
> it.

OK, private repositories, sure, I get it.  Can't share the content.

But the first I heard of this bug was muttering on #git, from your
customers, after you guys had to restore a repository EGit messed
up for him.  Its not like you don't know how to reach me, an email
would have been nice...

Also... regarding private projects... I have a tool in JGit that can
dump a scrubbed commit graph from a project.  I could have extended
it to dump trees too, but scrub file names and blob SHA-1s to hide
actual content.  Given that most corruption errors are structual,
it can be useful to dump the repository and examine what's going on.

I probably should get that improved and package as part of our
0.7.0 release.  At least then you can ask a user with a private
repository if you can share that structual dump with us.

 
> It was becoming very time consuming for us to help
> Windows EGit users over and over again when this would happen - it was
> easier to recommend users not use EGit until this issue was fixed.

I agree, that was a wise decision.  But a heads up a bit sooner by
email would have been appreciated by yours truely.

> (We
> would have been fscking the repos on receive, but there was a bug in
> Git when alternates were involved that made that not possible for us.)

Huh, that's double plus ungood.  Has this been reported/fixed
upstream in C Git yet?
 
-- 
Shawn.