488760 - Don't try to perform hg merges on build slaves

Status: RESOLVED FIXED

(Release Engineering :: General, defect, P2)

Description

[Chris Cooper [:coop] (he/him)]

Some check-ins and back-outs to mozilla/build/wince/tools over the past few days caused us to get into a state on some of the build slaves where hg was trying to perform merges against local file versions and failing. I know this is an issue on at least the Mac and Linux slaves.

e.g.: 
http://production-master.build.mozilla.org:8010/builders/Linux%20mozilla-1.9.1%20l10n/builds/1808/steps/shell_18/logs/stdio

We should never be trying to merge on slaves. catlee suggested adding an .hgrc entry on the slaves to always accept the new/incoming version, specifically the following stanza:

> [ui]
> merge: internal:other

I've tried this out on one of the affected slaves (moz2-linux-slave14) and it seemed to work.

If no one has any objections, I can try to push an .hgrc file out to all slaves next week. 

If there's anything else we'd want in an .hgrc file (since I don't believe we have one currently on any slaves), please comment below.

Comment 1

[Chris Cooper [:coop] (he/him)]

Hoping to get this pushed out to all the staging slaves today.

Comment 2

[Chris Cooper [:coop] (he/him)]

I found a graphlog entry in an existing .hgrc file on one of the slaves, so I added it to the .hgrc I was pushing out.

The current .hgrc file looks like:

> [ui]
> merge: internal:other
> 
> [extensions]
> hgext.graphlog =

Pushed out that .hgrc file to the following staging slaves:

# staging-master
moz2-darwin9-slave03
moz2-darwin9-slave04
moz2-darwin9-slave08
moz2-linux-slave03
moz2-linux-slave04
moz2-win32-slave03
moz2-win32-slave04
moz2-win32-slave21

# staging-1.9-master
fx-linux-1.9-slave1
fx-linux-1.9-slave03
staging-1.9-master
fx-mac-1.9-slave1
fx-win32-1.9-slave1
fx-win32-1.9-slave03

Comment 3

[Nick Thomas [:nthomas] (UTC+12)]

(In reply to comment #2)
> I found a graphlog entry in an existing .hgrc file on one of the slaves, so I
> added it to the .hgrc I was pushing out.

That was probably me doing some debugging with branches at some point, glad we can have it along at the same time.

Comment 4

[Chris Cooper [:coop] (he/him)]

I've pushed out the .hgrc file to all our production slaves and random boxes that might need it. Ask me sometime about how fun it was to do that for the windows slaves. :/

I have explicitly *not* pushed this out to the try slaves or talos slaves. Is this something we want for those two classes of slaves?

Comment 5

[Chris Cooper [:coop] (he/him)]

Alice says Talos slave don't even have hg on them yet, so no .hgrc required there.

Try slaves perform full clobbers for each new build, so the value of a no-merge .hgrc file would be minimal there.