https://pantsbuild.org/ logo
Join Slack
Powered by
<@U06A03HV1> I’m expecting <https://github.com/pan...
# announce
a
@witty-crayon-22786 I’m expecting https://github.com/pantsbuild/pants/pull/6503 to go green - PTAL when convenient 🙂
w
@average-vr-56795: i went ahead and merged this... assuming it's green, you should be able to reset-to and release-from that sha
a
Thanks for the notes update! Very helpful! :)
w
Do you want me to release this once we get master green?
a
I did the release already…
Didn’t realise master wasn’t green…
w
Ah. Yea, there was a high probability it would be. One issue that @fancy-queen-20734 is looking at. Not a big deal: we'll reply to the release thread when we have a fix
a
Cool 🙂 Thanks!
f
hm. the failed test on my PR didn’t fail on the release commit--but the same shard timed out
The job exceeded the maximum time limit for jobs, and has been terminated.
I restarted the shard
w
it appears to be deterministic. please take a look.
f
It looks like shard 1 started failing mostly deterministically on 
tests/python/pants_test/backend/jvm/subsystems:incomplete_custom_scala
after https://github.com/pantsbuild/pants/commit/906e4e5f4077cd1627eb5b5400f9cd7d5bbb8fab
I don’t have a local reproduction
my hypothesis is that something is being incorrectly cached
it might also be that the ordering of the tests is important
CI runs them in a different order than they are run locally when just invoking 
./pants test tests/…
Control-C when it hits during pool init has pretty annoying behavior
trying running the ci.sh that shard runs to see if that allows me to reproduce it locally
got a local failure, but it was because I didn’t setup the env correctly. 🤦‍♀️
argh. It didn’t fail locally again.
w
well, maybe rather than attempting to repro, we could revert that change?
ie, see whether a PR that reverts that change goes green. and then go from there
@fancy-queen-20734: do you want to do that, or should i?
f
I know the root cause now
w
ah, ok
f
locally, I’m getting scala-2.12.4
on ci, I’m seeing scala 2.12.2
w
ehhh? is something floating?
f
yeah
so, somehow 2.12.2 and 2.12.something later are interacting
and TermNames are not compatible between them
w
that's odd. i don't see anything floaty there.
in ./testprojects/src/scala/org/pantsbuild/testproject/custom_scala_platform/custom_212_scalatools.build
f
hm
w
where are you seeing the later version?
oh... i think i maybe see something?
f
I see the later version locally
and the older one in CI
w
in https://github.com/pantsbuild/pants/commit/906e4e5f4077cd1627eb5b5400f9cd7d5bbb8fab , 
def _tool_classpath
calls 
self._memoized_scalac_classpath
... ? naming wise that is confusing
f
my guess is that somehow both the later and earlier ones end up on the same classpath
w
and i wonder if 
def style_classpath
is ending up with something from the compiler classpath mixed in?
i haven't seen an "ah-hah" yet, but.
...that looks like just a naming issue. the memoized method should still be correct.
@fancy-queen-20734: regardless: consider talking with @gorgeous-alligator-94763 about reverting that one while we investigate.
f
Could the master branch’s cache entry be the problem?
I’m going to see if having a revert of that will no longer trigger the issue
w
Yep, thanks.
We've cleared the caches a bunch of times, so I doubt it's a cache.
f
k
w
blargh... still fails with it reverted. i wonder if a maven / artifactory cache is busted somewhere...? or if twitter artifactory has a good artifact, and maven central doesn't?
one thing to try would be just bumping the version there to some other 2.12 version.
f
yeah
hm
I could just bump the version in the test to 2.12.4--I wonder if the point of the test is to catch this particular error though
w
the test is supposed to succeed, afaict.
f
yeah
I think bumping the version should fix it. I’ll give that a go
Just to double check, I deleted the master cache entry and restarted the shard on the latest commit on master https://travis-ci.org/pantsbuild/pants/jobs/430365101
w
sounds good.
f
restarted shard didn’t fail with the error after I blew away the cache--instead it timed out
w
sigh
another timeout there, although it clearly made it past the previous failure. will merge on green
3 Views
Open in Slack
PreviousNext
Powered by