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

Re: ra_serf abort

From: Lieven Govaerts <svnlgo_at_mobsol.be>
Date: Fri, 02 May 2008 14:42:58 +0200

Daniel Shahaf wrote:
> Justin Erenkrantz wrote on Thu, 1 May 2008 at 10:10 -0700:
>> On Thu, May 1, 2008 at 10:04 AM, Daniel Shahaf <d.s_at_daniel.shahaf.co.il> wrote:
>>> Yes, except that I don't know whether the bug is in Serf or in Subversion.
>>> (The abort(), however, is in Serf code.)
>> You should try trunk - not 0.1.2. (Yes, I'm overdue to release the
>> next version.) -- justin
> I did, and it segfaulted in 'make check'. Lieven said it was a known
> issue.

To be clear, there are two known issues involved here. Known as in 'I
know about them, but that's about it'. The issue that started this
thread is an issue in ra_serf. In short, when you give a bucket with the
request body to serf, it will consume and release that bucket. When the
request has to be sent a second time (due to authentication required or
retry), that bucket can't be used again and must be recreated.
Practically this means that everytime we pass a request to serf we must
not pass a bucket with the body, but a callback that'll create the
bucket on demand, twice if needed.
There are a few locations in ra_serf where we still pass a body bucket,
this need to be converted to use the callback function.

The segfault in 'make check' of serf is purely serf related. The
segfault itself comes from the testsuite after a test failed. I'm not
sure if that test failure is really an issue in serf, but it should be
solved anyhow.

Hope this makes things clearer.


To unsubscribe, e-mail: dev-unsubscribe_at_subversion.tigris.org
For additional commands, e-mail: dev-help_at_subversion.tigris.org
Received on 2008-05-02 14:43:17 CEST

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

This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.