[users at bb.net] Collapsing Builds Per Branch?

[Neil Gilmore]

In 0.9.3 at least (and the current code doesn't appear to be different 
for this), we modified to code in buildrequest.py to ignore revision 
when testing for collapsibility. Otherwise, every commit was causing a 
non-collapsible build to be scheduled. We already set treeStableTimer, 
but that didn't help in this case. I do not recall that we had such a 
modification on 0.8, but I wouldn't count on my memory there.

To quote process/buildrequest.py, down around line 275:
for c, selfSS in iteritems(selfSources):

otherSS = otherSources[c]

if selfSS['revision'] != otherSS['revision']:

defer.returnValue(False)

return

Also Chris, in 0.9, cancelling the queue from the UI stops any currently 
executing build.

Neil Gilmore
ngilmore at grammatech.com

On 5/15/2017 10:11 PM, Charles Lepple wrote:
> On May 15, 2017, at 12:29 PM, Chris Spencer wrote:
>>
>> It should also cancel a running build for a branch if there's a newer 
>> commit pending.
>
> I don't think that automatically cancelling a build once it starts has 
> ever been part of the base scheduler classes (at least, in 0.8). You 
> might be able to do this by hitting the web status interface from 
> another scheduler, though it might make sense to have a heuristic for 
> allowing a build to run to completion if it is nearly done anyway.
>
> The collapse options are for queued builds (modulo the 0.9.x bug that 
> Pierre mentioned), and the treeStableTimer can be used to wait for 
> several commits to arrive, in order to give them a chance to be collapsed.
>
>
>
> _______________________________________________
> users mailing list
> users at buildbot.net
> https://lists.buildbot.net/mailman/listinfo/users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.buildbot.net/pipermail/users/attachments/20170516/9d3913bc/attachment.html>