1254804 - mozregression doesn't properly visit inbound builds.

Status: RESOLVED FIXED

(Testing :: mozregression, defect)

Description

[Jeff Muizelaar [:jrmuizel]]

In bug 1251868 we ran into this weird log:
https://bug1251868.bmoattachments.org/attachment.cgi?id=8727934

which caused us to not properly bisect the inbound builds.

Comment 1

[Julien Pagès (:parkouss)]

From what I can see, we ends up with the following last pushlog:

https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=c1e0d1890cfee9d86c8d566b0490053f21e0afc6&tochange=97cf677ee66802809808a3e61a0ccb89542ca54e

And this looks like a merge, but without information about it - so mozregressio does not know where these commits comes from and just stop. Indeed, they are on inbound:

https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?tochange=97cf677ee668&fromchange=9ecc9682c3a3

So you can manually ask mozregression to bisect that:

> mozregression --good 9ecc9682c3a3 --bad 97cf677ee668 --repo inbound

Eventually, we could make mozregression smarter, trying to bisect on mozilla-inbound automatically if we don't find any merge info and that we have multiple commits with different authors. (So it should apply the above logic automatically)

Comment 2

[Julien Pagès (:parkouss)]

Attachment: auto switch on m-i

This should be good - it works for the current example at least.

Comment 3

[William Lachance (:wlach)]

Comment on attachment 8728324 [details] [review]
auto switch on m-i

Thanks!! One comment in the PR.

Comment 4

[Julien Pagès (:parkouss)]

Thanks Will - and sorry for the dely here. :) I fixed the message as you suggested. I'm going to create a patch release with this fix now.

landed in https://github.com/mozilla/mozregression/commit/2a886a5a79f2afcef2c073ca340006f7a8fe27c6.