Bug 39282 – Slave lost shouldn't be recognized as build failed

• Status: RESOLVED
• Resolution: FIXED
• Component: Tools / Tests

[Csaba Osztrogonác]

Unfortunately slave lost is recognized as build failed now,
and sherrifbot bother innocent folks with false positive alarms
on IRC and add comments into bugzilla.

I propose sherrifbot should handle "slave lost" as "build successful".

[Csaba Osztrogonác]

Created attachment 56355 [details]
proposed patch

[Eric Seidel (no email)]

I don't think this is the right fix.  But it's an interesting hack.

For one, we would need to unit test this.

Secondly, I think the right fix would be to crawl back a revision and use that as the current red/green status instead of just assuming "slave-lost" means green.  Although maybe we should just treat lost slaves as green?  Not sure.

[Eric Seidel (no email)]

Comment on attachment 56355 [details]
proposed patch

Please add a FIXME comment.  Something like:

# FIXME: We treat slave lost as green even though it is not to work
# around the Qts bot being on a broken internet connection.
# The real fix is https://bugs.webkit.org/show_bug.cgi?id=37099

But this solution is pragmatic and we should do it.

[Csaba Osztrogonác]

Landed in http://trac.webkit.org/changeset/60559

[Csaba Osztrogonác]

(In reply to comment #4)
> Landed in http://trac.webkit.org/changeset/60559
And rolled-out by http://trac.webkit.org/changeset/60560
because I forgot about unit test and broke the whole world.

Sorry. Fix is coming soon.

[Csaba Osztrogonác]

Created attachment 57659 [details]
Patch

[Csaba Osztrogonác]

(In reply to comment #6)
> Created an attachment (id=57659) [details]
> Patch

Here is the fixed patch with unit test for slave lost.
I have to add a "not not", because False or None == None.
Or is there any simple way to fix this problem?

Error message why I had to roll-out my earlier patch:

Traceback (most recent call last):
  File "/Volumes/Data/WebKit-BuildSlave/snowleopard-intel-leaks/build/WebKitTools/Scripts/webkitpy/common/net/buildbot_unittest.py", line 218, in test_status_parsing
    self.assertEquals(builder[key], expected_value, ("Builder %d parse failure for key: %s: Actual='%s' Expected='%s'" % (x, key, builder[key], expected_value)))
AssertionError: Builder 2 parse failure for key: is_green: Actual='None' Expected='False'

[Eric Seidel (no email)]

Comment on attachment 57659 [details]
Patch

LGTM.  This seems like the right fix.

[Csaba Osztrogonác]

(In reply to comment #8)
> (From update of attachment 57659 [details])
> LGTM.  This seems like the right fix.
Landed in http://trac.webkit.org/changeset/60575 .