> From: mikem@redhat.com
> On 10/21/2013 10:32 AM, John.Florian@dart.biz wrote:
> > koji regen-repo f19-build
> > Regenerating repo for tag f19-build
> > Watching tasks (this may be safely interrupted)...
> > 8 newRepo (f19-build): free
> > SysCallError: (-1, 'Unexpected EOF')
> I'm seeing errors like this on F19 myself. They are intermittent and
> normally retried by koji, but the cli calls that watch tasks drop
> authentication before they start watching.
> You can work around this by enabling anon_retry in your koji config
> (e.g. in ~/.koji/config).

Okay, I've added "anon_retry=true" there.

> The good news is that this appears to be simply the task watcher failing
> on the client side. Your actual tasks may well be working.

Ah! That may help explain what I observed after posting that mail.  While trying to figure out why the newRepo call was failing I learned that this particular job gets handed off to a builder and then got more logging going for kojid where I learned that my builder was enabled but not ready.  While trying to figure out why it wasn't ready, it suddenly became ready and somewhere in that vicinity of time I also realized that the newRepo had succeeded.  I never felt that I'd made any change to explain the success, but was fiddling quite a bit trying to make forward progess.

> You can also always run koji watch-task $N to restart the watch if it dies.

Good to know.  Thanks!

> > I've also tried doing a scratch build and see much of the same ...
> >
> > koji build --scratch --wait f19
> > /pub/fedora/mdct/19/SRPMS/plant-launchers-1.3-1.fc19.src.rpm
> > Uploading srpm:
> > /pub/fedora/mdct/19/SRPMS/plant-launchers-1.3-1.fc19.src.rpm
> > [====================================] 100% 00:00:25  11.10 KiB   449.99
> > B/sec
> > Created task: 10
> > Task info:
> > Watching tasks (this may be safely interrupted)...
> > 10 build (f19, plant-launchers-1.3-1.fc19.src.rpm): free
> > SysCallError: (-1, 'Unexpected EOF')
> >
> >
> > I'm not seeing anything helpful in the logs and since I'm such a koji newb
> > I'm not sure what to make of this problem.  Any advice would be greatly
> > appreciated.
> I believe this is happening client side for you. If you run your
> commands with the --debug global option, you'll probably see a traceback
> similar to this:
>   File "/usr/lib/python2.7/site-packages/koji/__init__.py", line 1925,
> in _callMethod
>     return self._sendCall(handler, headers, request)
>   File "/usr/lib/python2.7/site-packages/koji/__init__.py", line 1836,
> in _sendCall
>     return self._sendOneCall(handler, headers, request)
>   File "/usr/lib/python2.7/site-packages/koji/__init__.py", line 1856,
> in _sendOneCall
>     response = cnx.getresponse()
>   File "/usr/lib64/python2.7/httplib.py", line 1045, in getresponse
>     response.begin()
>   File "/usr/lib64/python2.7/httplib.py", line 409, in begin
>     version, status, reason = self._read_status()
>   File "/usr/lib64/python2.7/httplib.py", line 365, in _read_status
>     line = self.fp.readline(_MAXLINE + 1)
>   File "/usr/lib64/python2.7/socket.py", line 476, in readline
>     data = self._sock.recv(self._rbufsize)
>   File "/usr/lib/python2.7/site-packages/koji/ssl/SSLConnection.py",
> line 140, in recv
>     return con.recv(bufsize, flags)
> SysCallError: (-1, 'Unexpected EOF')
> I'm not yet sure what the cause of this is, but I suspect some behavior
> change in the underlying libs.

That indeed looks familiar.  Unlike the newRepo, I cannot make a build succeed (i.e., no binary rpm) despite lots of waiting.  Sounds like Koki atop F19 isn't ready for prime time yet, especially for a newb as myself.  Can you recommend a better Fedora release?  I've got much of the setup codified in puppet so it shouldn't be too much a PITA to rebuild.

John Florian