On Tue, 2009-09-22 at 08:20 +0200, Lieven Govaerts wrote:
> On Tue, Sep 22, 2009 at 1:32 AM, Julian Foad <julianfoad_at_btopenworld.com> wrote:
> > On Tue, 2009-09-22 at 00:27 +0100, Julian Foad wrote:
> >> On Sun, 2009-09-20 at 22:04 +0200, Lieven Govaerts wrote:
> >> > On Thu, Sep 17, 2009 at 11:58 PM, Lieven Govaerts <svnlgo_at_mobsol.be> wrote:
> >> > > This particular script doesn't trigger a hang on my machine, but I get
> >> > > a reproducible hang in svnsync with serf 0.3 now which I've reported a
> >> > > while ago as issue 3424. Big chance that it's the same issue as you
> >> > > (and others) are seeing.
> >> >
> >> > I've fixed the issue I've been seeing, but it's related to opening a
> >> > child process. So, unless you're using an external diff program, this
> >> > fix can't solve your problem.
> >> >
> >> > Next time it hangs, could you see with wireshark (or another tcp dump
> >> > program) if you get ZeroWindow packets? The issue I solved caused serf
> >> > to hang in apr_pollset_poll, with no sockets to listen on. Maybe it
> >> > can be triggered in other ways.
> >> When mine hangs, my svn was sending TCP ZeroWindow packets to
> >> svn.collab.net every second or two for the last 15 seconds before the
> >> hang. From when it hangs, there are no further packets of any kind
> >> either sent or received between my machine and svn.collab.net. I waited
> >> for at least a minute before stopping the capture. I can send you the
> >> trace if it helps.
> No that's okay, it's the same symptom as I fixed last weekend,
> question is if it's the same rootcause.
> I've read on IRC that you're using an external diff3 program. Can you
> try with serf @ http://serf.googlecode.com/svn/branches/0.3.x to see
> if that fixes your problem?
I tried my test again with Serf 0.3.x_at_1275 (with your debugging patch
applied), and it worked properly (didn't hang).
Received on 2009-09-22 16:08:26 CEST