> On Nov 1, 2016, at 2:22 PM, Mojca Miklavec <mo...@macports.org> wrote:
> 
>> On 1 November 2016 at 18:07, Lawrence Velázquez wrote:
>> 
>>> On Nov 1, 2016, at 7:00 AM, Mojca Miklavec wrote:
>>> 
>>> Why not:
>>> - wait until all the slaves have something non-trivial to do (make
>>> sure the queue is not empty)
>>> - do a bulk commit
>>> - cancel the job manually
>> 
>> That won't prevent the buildmaster from receiving 20,000 changes,
>> processing ~100,000 build requests, and scheduling ~100,000 builds.
> 
> No, that's not true. 100.000(?) build request would only be scheduled
> if we fail to cancel the build on the portwatcher.

I don't think this is accurate. Builds would not yet have been created,
but BuildRequests would have been scheduled.

But I think you're right, for the wrong reasons. I forgot that the port
scheduler has a tree stable timer now; a single BuildSet would be
created for all 20,000 Changes.

> If we cancel the job on the portwatcher before
> it gets processed, there will be absolutely zero penalty (except that
> the failcache would probably be erased).

Just because builds are not actually happening yet does not mean there
isn't a memory penalty from the queued requests.

In any case, the proposed set of commits is utterly unnecessary and is
not worth any of this fuss.

vq
_______________________________________________
macports-dev mailing list
macports-dev@lists.macosforge.org
https://lists.macosforge.org/mailman/listinfo/macports-dev

Reply via email to