On 08/22/2014 01:48 PM, Richard Hughes wrote:

So builds that have waited more than 6 hours are just killed? That's going to be really difficult to deal with, when repos like codeblock/ghc-78-everything send thousands of builds which take half a day to clear...

No. Build which are *running* for 6 hours. This is time from moment when they have state set to running.

For example right now it is build: http://copr.fedoraproject.org/coprs/mvala/saske/build/28014/ which is running for several days and is blocking one of the builders.

To repeat it - time spent in queue does not matter. Time spent during actual build time is what matters.

On 08/22/2014 07:48 AM, Richard Hughes wrote:

when repos like codeblock/ghc-78-everything send thousands of builds which take half a day to clear...

I'm sorry for clogging things up with that set of builds. I will do one Fedora version at a time instead of two in the future, if I do that again.

I did try cancelling one of the versions (f21, both architectures) but the builds were already queued and it was too late, so I re-added it.

I hope I didn't cause too much trouble - we (haskell-sig) are just trying to see where we stand so far in terms of getting ghc 7.8 in Fedora, and we thought Copr would be a good way to do all of the necessary builds, especially with its monitor page, we can very easily see all of the failing packages, etc. But I probably should've asked first. Sorry again.

-Ricky