1227473 - Something up with the b2g bumper gaia update from 15 November

Status: RESOLVED WONTFIX

(Developer Services :: General, task)

Description

[Pete Moore [:pmoore][:pete]]

See https://hg.mozilla.org/integration/fx-team/file/4916558f00d5/b2g/config/gaia.json

git sha is correct: https://github.com/mozilla-b2g/gaia/commit/28d63cf3bdc4417f7ad8cab2230f096bf9f6d3b5

hg sha is incorrect: https://hg.mozilla.org/integration/gaia-central/file/f1fbe09c3fd9e47b2ca2709b471b2c9cfd7a733f

This caused: https://public-artifacts.taskcluster.net/fg0gQHrDQ3SACttGE9no4Q/0/public/logs/live_backing.log

[taskcluster-vcs] 1 run start : (cwd: /home/worker/gaia) hg pull -r f1fbe09c3fd9e47b2ca2709b471b2c9cfd7a733f https://hg.mozilla.org/integration/gaia-central
pulling from https://hg.mozilla.org/integration/gaia-central
abort: unknown revision 'f1fbe09c3fd9e47b2ca2709b471b2c9cfd7a733f'!
[taskcluster-vcs:error] run end (with error) NOT RETRYING!: hg pull -r f1fbe09c3fd9e47b2ca2709b471b2c9cfd7a733f https://hg.mozilla.org/integration/gaia-central

Comment 1

[Pete Moore [:pmoore][:pete]]

Looks like the real hg sha is 1b1d1381eadbfd3de6c89bb39f28f2aa396183a8 and not f1fbe09c3fd9e47b2ca2709b471b2c9cfd7a733f.

See: https://hg.mozilla.org/integration/gaia-central/rev/1b1d1381eadb

I've no idea if:

a) originally it was f1fbe09c3fd9e47b2ca2709b471b2c9cfd7a733f but then later the git->hg conversion process changed and overwrote it
b) there is a bug in bumper bot such that it published the wrong SHA
c) there was a force-push or repo history update outside of b2g bumper/vcssync

Comment 2

[Pete Moore [:pmoore][:pete]]

Note, this may or may not be related to the commit message being truncated in hg:

In git: "Bug 1222204 - Add Persian (fa) to master and 2.5 builds, r=stas, a=mahe"
In hg:  "Bug 1222204 - Add Persian (fa) to master and 2.5 builds, r=stas"

Comment 3

[Pete Moore [:pmoore][:pete]]

Might also be worth checking what legacy mapper thinks the mappings between git and hg are for this commit... Not sure the url for the mappings file.

Handing over to RelEng to take from here. =)

Comment 4

[Hal Wine [:hwine] use NI!]

:pmoore - thanks for all the data. 

Questions: 
 - What is the impact currently?
 - If none now, any issue besides ability to bisect?

Comment 5

[Pete Moore [:pmoore][:pete]]

I'm not aware of any impact this had other than jmaher raising the matter in #taskcluster channel some hours ago.

However, if there are incorrect commit shas being written in gaia.json, there could be problems coming our way...

Comment 6

[Joel Maher ( :jmaher ) (UTC -8)]

I am trying to retrigger to find where issues were introduced and this is proving to be impossible if we cannot retrigger a job.  Is there a way to fix this or hack around it?

Comment 7

[Hal Wine [:hwine] use NI!]

This is likely fallout from bug 1224628, investigating.

Comment 8

[Hal Wine [:hwine] use NI!]

Workaround: rebuild the gecko revision-of-interest after updating the b2g/config/gaia.json file to reference the correct (current) hg hash for the git revision. Ask for help in #releng if you need help obtaining the correct hash.

The fix to bug 1224628 caused hashes to be changed on the hg copies of gaia branches master and v2.5. There was a window of time when jobs would have had a no-longer-valid hash plugged into their gaia.json file.

Attempting any further operation on the original job (such as re-triggering a test) which attempts to pull the gaia source from the hg copies will fail. 

Doing a complete rebuild with the updated b2g/config/gaia.json file will use the identical source code as the original build. Only the hash has changed.