1146379 - hgtool should avoid pulling if it already has a revision

Status: RESOLVED FIXED

(Release Engineering :: General, defect)

Description

[Chris AtLee [:catlee]]

Most cases in automation specify the revision they want to checkout to hgtool. Right now it's always does a pull prior to updating to the revision.

We could save on some load by checking first to see if we have the revision locally prior to contacting the remote.

Comment 1

[Gregory Szorc [:gps]]

Great idea!

Comment 2

[Chris AtLee [:catlee]]

Attachment: check to see if we have a revision before pulling

Comment 3

[Jordan Lund (:jlund)]

Comment on attachment 8583076 [details] [review]
check to see if we have a revision before pulling

backed out on suspected bustage for Try:
remote:   https://hg.mozilla.org/build/tools/rev/3e1b71cac152

failed builds started rolling in around the time of this deployment complaining about Revisions. I suspect this is because Try is a special snowflake that calls hgtool differently (with --revision for example)

log examples:
https://treeherder.mozilla.org/logviewer.html#?job_id=5948205&repo=try
https://treeherder.mozilla.org/logviewer.html#?job_id=5953389&repo=try

Comment 4

[Jordan Lund (:jlund)]

hrm, maybe this is not the cause of failed jobs but treehereder is highlighting the errors:

15:27:00    ERROR -  abort: unknown revision 'af9c9ff307d0'!
15:27:00    ERROR -  Automation Error: hg not responding

but it looks like it recovers after:

15:27:30     INFO -  command: output:
15:27:30     INFO -  af9c9ff307d0
15:27:30     INFO -  Got revision af9c9ff307d0
15:27:30     INFO - Return code: 0

Comment 5

[Jordan Lund (:jlund)]

yes it looks like all green builds are getting highlighted in try with:

"abort: unknown revision '$SOME_REV'!" 

we should probably just surpress/catch that output. Sounds like the exception subprocess.CalledProcessError is not enough

Comment 6

[Chris AtLee [:catlee]]

Attachment: check to see if we have a revision before pulling

Fixed up the logging a bit. It now looks like this on success:

...
checking to see if 2d6cc215dc64 exists in /home/catlee/test/buildbotcustom
command: START
command: hg log -r 2d6cc215dc64 --template {node}
command: cwd: /home/catlee/test/buildbotcustom
command: env: {'HGPLAIN': '1'}
command: END (0.25s elapsed)

command: output:
2d6cc215dc6480808296d386cf8b97a04ef4785e
2d6cc215dc64 exists in /home/catlee/test/buildbotcustom
skipping pull since we already have 2d6cc215dc64
...

and on failure:

...
checking to see if 2d6cc215dc65 exists in /home/catlee/test/buildbotcustom
command: START
command: hg log -r 2d6cc215dc65 --template {node}
command: cwd: /home/catlee/test/buildbotcustom
command: env: {'HGPLAIN': '1'}
2d6cc215dc65 does not exist in /home/catlee/test/buildbotcustom
...