I tested the patch by running the patched version under the bigger software
I'm using and my own debugging statements in pycurl and tornado. So
there're no unit tests for the bug, sorry. As to the bug report, since I
couldn't find an issue tracker for pycurl, I posted all relevant info on
this list.

On 18 April 2013 20:37, Oleg Pudeyev <oleg+pycurl_at_bsdpower.com> wrote:

> Thanks for the patch.
>
> Is there a bug report or code to reproduce the bug?
>
> Oleg
>
> On Thu, 4 Apr 2013 15:40:24 +0400
> Yuri Ushakov <yuri.ushakov_at_gmail.com> wrote:
>
> > http://dl.dropbox.com/u/8792466/pycurl.c.patch
> >
> > When really big timeouts were specified, do_multi_timeout() would
> > return negative values to Python code. Libcurl takes long and handles
> > it well, whereas pycurl corrupted the value.
> >
> > That made tornado (http://www.tornadoweb.org) go into an essentially
> > infinite loop and stop serving incoming http requests, eating up all
> > CPU time.
> >
> > Tested on Python 2.6.5 and Tornado 2.0.0.
>
>

_______________________________________________
http://cool.haxx.se/cgi-bin/mailman/listinfo/curl-and-python
Received on 2013-04-18