[webkit-dev] Rolling out a patch requires a justification beyond a test failure in downstream projects

[Maciej Stachowiak]

On Dec 11, 2012, at 7:34 PM, Ryosuke Niwa <rniwa at webkit.org> wrote:

> On Tue, Dec 11, 2012 at 7:20 PM, Elliott Sprehn <esprehn at chromium.org> wrote:
> Do you have an example of when this has occurred? It's good to have examples if we want to prevent this in the future.
> 
> Yes. I'd rather not publicly humiliate someone on webkit-dev so I'll send you a bug URL in private.
> 
> In this particular incident, a WebKit patch was rolled out due to a Chromium UI test failure. The person who rolled out the patch didn't communicate any information on the original bug from which the patch was landed. On the bug where the rollout was made, the person left links to Chromium WebKit roll patches but without any information regarding tests that failed. The only reason I could follow his comments is because I used to work on Chromium. The patch was subsequently rolled out in less than 20 minutes from the time the person first provided any information about the test failure at all.
> 
> To make it even worse, the roll out was speculative, and the patch was found innocent of causing the failure. The person promised to re-land the patch by Monday and never came back to the bug. (Now you know where my previous email came from).

In addition to the wrongnesses you pointed out, the combination of a speculative rollout for a downstream test failure (where clearly you should have the ability to test if the patch is really to blame) seems even more extra wrong.

I agree with not embarrassing people needlessly on webkit-dev, but someone should probably have a direct conversation about all this with the individual responsible.

 - Maciej

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.webkit.org/pipermail/webkit-dev/attachments/20121211/0552c6d8/attachment.html>