savannah-hackers-public
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Savannah-hackers-public] Savannah bzr server errors out


From: Martin Pool
Subject: Re: [Savannah-hackers-public] Savannah bzr server errors out
Date: Mon, 23 Jan 2012 13:03:31 +1100

On 21 January 2012 20:26, Eli Zaretskii <address@hidden> wrote:
>
> > From: Bake Timmons <address@hidden>
> > Date: Fri, 20 Jan 2012 17:52:58 -0500
> >
> > I'm unable to upgrade my emacs trunk repo past certain revisions (106781
> > and 106890).  Rather than download a whole bzr repo for a third time, I
> > am reporting this problem and will just use a git repo until I can use
> > bzr again.
> >
> > What the 'bzr pull' failures have in common is that they occur after
> > about 2-2.5 hours on my 28.8 kbps dialup connection (a little over 20MB
> > of data transferred).  The failures happen using bzr, nosmart+bzr, and
> > http protocols and under the bzr client versions 2.5b4 and 2.5b5.
> > Running 'bzr check' showed nothing wrong with the repo, in the case of
> > emacs trunk version 106781 and in the case of version 106890.
> > [...]
> >   File "/usr/lib/python2.7/dist-packages/bzrlib/smart/message.py", line 
> > 286, in _read_more
> >     "Unexpected end of message. "
> > ConnectionReset: Connection closed: Unexpected end of message. Please check 
> > connectivity and permissions, and report a bug if problems persist.
>
> It's a "feature": the savannah bzr server has a 2-hour timeout, so it
> simply closes the connection after that.
>
> The timeout was introduced because otherwise zombie instances of the
> bzr server would be left after broken connections and gradually bring
> bzr.savannah to its knees.
>
> I'm told that bzr 2.5, to be released in a few weeks, will have
> built-in features that will allow to remove the timeout.  I hope once
> bzr 2.5 is out, we could have it installed on savannah and fix this
> problem.
>
> Until then, I'm afraid your only choice would be to use a faster link.
>
> Martin, is there any other workaround?  For that matter, did I
> describe the issue correctly?

Hi, Eli,

It looks like it was actually running at full wire speed (3.1kB/s!).
I wonder if there really should have been 20MB of updates to download
- a -Dhpss run would help tell that - but it seems possible a fairly
stale emacs checkout would need that many.

You could install 2.5b5 now - it is at release stability now and it
will be useful to get some feedback on whether it helps people like
Bake.  Using 2.5b5 on the client may help too.

--
Martin



reply via email to

[Prev in Thread] Current Thread [Next in Thread]