Hi Ruben! I had a few tickets with Oliver upstream. There were a few cornercases that my (quite large) setup triggered quickly. Seems that's all fixed by now. These included:
* Under some circumstances, laminard didn't accept any further web
requests when a client shut down the connection kind of at the
wrong point. That was an issue with capnproto
(https://github.com/capnproto/capnproto/pull/1407), which has no
new release yet after that was merged. I hope Tom Lee will package
capnproto quickly with the next release. Right now, Debian's 0.7
version is quite dated...
* Due to a timing/locking bug, fetching (or following) logs didn't
work properly when there was new log data received from a running
build script while laminard was sending already received data to
a web client. (Quite easy to trigger with jobs that produce lots
of logging, where downloading initial megabytes of logs may take
some time.)
* Also, laminarc can now reorder already queued jobs, so that when
there are many jobs queued (in my case: usually well over 1000,
running for about a week to ten days to work these down), you can
give priority to some. (Usual scenario is to trigger a new run
for something that you think you just fixed and you don't want to
wait a week for the result to show up.)
So it would be nice to update the laminard/laminarc packages to
current versions, and possibly trigger a rebuild when there's a new
capnproto available? (I'm not sure how much of their *headers*
actually end up in the binary...)
Thanks,
Jan-Benedict
--
signature.asc
Description: PGP signature
