[darcs-users] Another possible bug in URL.waitNextUrl: curl_multi_perform?

Eric Kow kowey at darcs.net
Fri Apr 2 15:43:45 UTC 2010

Hi everybody,

We've been receiving quite a few bug reports with an error message like
the above.

  Another possible bug in URL.waitNextUrl:  curl_multi_perform() - no
  running handles

First, does anybody else see these error messages?  If so, please file
a ticket on bugs at darcs.net including if possible

 - darcs version
 - libcurl version
 - flags for compiling darcs (particularly, if you're using -fcurl-pipelining)
 - if you can reproduce this at all, and provide --debug-http output

Second, below are the current reports I'm tracking.  Some of these may
all different bugs with the same symptoms but it's hard to tell.  If
anything jumps out at you, please shout :-)

=========  =======  ===============   =====  ======  ============
Origin     Date     Reporter          Darcs  Curl    Notes
=========  =======  ===============   =====  ======  ============
issue1368  2009-02  Zooko (Tahoe BB)  2.2.1          intermittent
issue1420  2009-04  Zooko (Tahoe BB)  2.2.0          intermittent
email      2009-06  Corinna           2.2.0          intermittent
issue1489  2009-07  Larisse           2.2.0          permissions?
email (*)  2009-10  GHC BB            -      
issue1770  2010-03  Trent             2.3.0  7.20.0  intermittent; squid proxy?
issue1808  2010-04  Stelian           2.4.0  7.20.0  systematic! pipelining! curl bug?
=========  =======  ===============   =====  ======  ==========

- GHC buildbot http://www.haskell.org/pipermail/cvs-ghc/2009-October.txt

- Larisse: Fedora 10, may be curl 7.19.4?

- Larisse said that he was just trying to get into a repo with
  insufficient permissions and that sudo darcs get 'solved' it.
  Seems quite confusing

- Stelian suspects "faulty address ordering/selection algorithm in curl"
  (he has seen similar problems with git!), but given that his is the
  only systematic instance of this bug, I suspect that he's found a
  different problem.  He's on the case, probably trying to debug this
  for upstream (curl).  Hopefully his insights will trickle down...


Eric Kow <http://www.nltg.brighton.ac.uk/home/Eric.Kow>
PGP Key ID: 08AC04F9
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: Digital signature
URL: <http://lists.osuosl.org/pipermail/darcs-users/attachments/20100402/9887d179/attachment.pgp>

More information about the darcs-users mailing list