hg push aborted message is displayed but push was still preformed

[Matt Mackall]

On Fri, 2014-04-25 at 11:27 +0100, Keren Asulin wrote:
> >
> > Yep. It's an occasional glitch that the Linux NFS client should paper
> > over.
> >
> > We've been using mercurial for over 2 years this only started in the last
> > month, also for an occasional glitch it's somewhat frequent and can last
> > for a while.
> > Are there any more details you can give me about this issue? anything we
> > can change/upgrade that may make this go away?
> >
> > Some more important data:
> We upgraded mercurial from 1.6 to 2.8 about a month before this problem
> started (we also migrated the server from debian squeeze to ubuntu 12.4) -
> this problem did not occur in mercurial 1.6 or it was too infrequent to
> notice.

Ok, Ubuntu 12.04 might be too new to hit the NFS issue. My second theory
is that you're running into a number of processes ulimit or similar. On
Unix, fork() is one of the few system calls that will actually return
EAGAIN in normal usage. This correlates well with it occurring during
hook execution.

Another possibility is that your Jenkins hook is an in-process Python
hook (dis-recommended and unsupported) and is doing its own non-blocking
or timeout-enabled socket I/O without appropriate error handling.

But let me be 100% clear about this point: whatever the issue is, it is
not a Mercurial issue. The message "Resource temporarily unavailable" is
coming from the kernel on your server and as such is as "real" as can be
and isn't something Mercurial can ignore.

-- 
Mathematics is the supreme nostalgia of our time.