Re: Proposal to change the semantics of DONTBUILD
On Tuesday 2013-01-15 13:20 -0800, Gregory Szorc wrote: This seems to make sense. My only concern is if there is a scenario where you absolutely need to push without incurring a build (think merge commit where you don't have control over the previous commits). I'm not sure why we'd do that, so I'm inclined to believe that such a scenario does not exist. I haven't heard of an absolutely need to not build scenario. We sometimes do extra builds for various reasons. The point of DONTBUILD is to save resources; occasionally failing to save resources when we could have done so is much less of a problem than not doing things that we really need to do. So Ehsan's proposal sounds great. -David -- 턞 L. David Baron http://dbaron.org/ 턂 턢 Mozilla http://www.mozilla.org/ 턂 ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Proposal to change the semantics of DONTBUILD
Hi all, Currently, DONTBUILD only takes affect if it's set on the tip of the changesets that you push. This can cause problems when merging between different trees if the target tree is a full subset of the origin tree (i.e., fast-forward merges in git terminology) when the top-most changeset is DONTBUILD, since it means that the entire push will not get builds and tests. Over in bug 815215, I've proposed that we change DONTBUILD to have per-changeset semantics, meaning that for each patch that does not require builds and tests, you should add DONTBUILD to your commit message. With this change, buildbot will look at all changesets in your push and will only skip builds if all of them are DONTBUILD. What this means to you as a person who merges trees is that you'll get sane semantics by default. If you're a person who writes DONTBUILD changes, this means that you should now specify it on all individual patches (which in practice only makes a difference if you push more than one DONTBUILD changeset at a time to let's say inbound.) Does anybody object to this proposal? Cheers, -- Ehsan http://ehsanakhgari.org/ ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Proposal to change the semantics of DONTBUILD
On 1/15/13 1:11 PM, Ehsan Akhgari wrote: Hi all, Currently, DONTBUILD only takes affect if it's set on the tip of the changesets that you push. This can cause problems when merging between different trees if the target tree is a full subset of the origin tree (i.e., fast-forward merges in git terminology) when the top-most changeset is DONTBUILD, since it means that the entire push will not get builds and tests. Over in bug 815215, I've proposed that we change DONTBUILD to have per-changeset semantics, meaning that for each patch that does not require builds and tests, you should add DONTBUILD to your commit message. With this change, buildbot will look at all changesets in your push and will only skip builds if all of them are DONTBUILD. What this means to you as a person who merges trees is that you'll get sane semantics by default. If you're a person who writes DONTBUILD changes, this means that you should now specify it on all individual patches (which in practice only makes a difference if you push more than one DONTBUILD changeset at a time to let's say inbound.) This seems to make sense. My only concern is if there is a scenario where you absolutely need to push without incurring a build (think merge commit where you don't have control over the previous commits). I'm not sure why we'd do that, so I'm inclined to believe that such a scenario does not exist. ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Proposal to change the semantics of DONTBUILD
On 01/15/2013 01:20 PM, Gregory Szorc wrote: This seems to make sense. My only concern is if there is a scenario where you absolutely need to push without incurring a build (think merge commit where you don't have control over the previous commits). I'm not sure why we'd do that, so I'm inclined to believe that such a scenario does not exist. In that scenario (if it exists) you could always use the build API (and the shortcut red-stopsign buttons on TBPL) to effectively get what you want. +1 to ehsan's proposal. ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Proposal to change the semantics of DONTBUILD
On 15/01/2013 23:52, Daniel Holbert wrote: In that scenario (if it exists) you could always use the build API (and the shortcut red-stopsign buttons on TBPL) to effectively get what you want. Killing running builds would mean needing to clobber all machines for all platforms for that tree (until we have bug 658934), which would possibly outweigh the benefit (depending on build vs test machine load at the time). However I think this scenario is sufficiently uncommon that it needn't block Ehsan's proposal :-) Best wishes, Ed ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform