[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: Commit error using 1.8.0 final

From: Ben Reser <ben_at_reser.org>
Date: Tue, 18 Jun 2013 07:55:38 +0200

On Tue, Jun 18, 2013 at 1:33 AM, Mark Phippard <markphip_at_gmail.com> wrote:
> I sent Lieven a tcpdump. Hopefully that will help. In the meantime I
> found another tool that analyzes HTTP traffic. The last request
> before it fails is a DELETE request that the server responds with a
> 204. Would that be unexpected by Serf?
> The URL for the DELETE request is:
> http://subclipse.tigris.org/svn/subclipse/!svn/act/b61b1e79-ba56-41b3-8b07-8649146e8212

The 204 shouldn't be odd, that just means a response without content.

However, the fact that you have requests happening after a DELETE
request seems odd. This particular DELETE is a activity URL, which
means it is removing the transaction on the server. If for some
reason we're sending this DELETE before we finish sending the other
requests that need to be sent for the commit, that would obviously be
very bad.

Are you sure the DELETE request is not happening after the failure?
Because it would be normal for the client to send DELETE after it
encountered some problem.

> Right before that (maybe Serf has multiple connections going) is a
> CHECKOUT that gets a 201. I notice that one says it was Closed by
> peer, so that might be the problem. Here is the info I can get out of
> this tool:

Possible the client/serf thinks the checkout fails due to whatever is
causing that "Closed by peer" you're seeing, which would trigger error
and the DELETE. Hard to say what any traffic after the DELETE is
without knowing more than you posted on the list.
Received on 2013-06-18 07:56:25 CEST

This is an archived mail posted to the Subversion Dev mailing list.