https://pantsbuild.org/ logo
Join Slack
Powered by
travis ci appears to not be showing a yellow in-pr...
# general
a
travis ci appears to not be showing a yellow in-progress circle if the build has not started yet on two of my open PRs -- the builds are still queued, but i wanted to know if anyone else was seeing this (i think it's a cool feature if it's a feature)
e
travis or github? github shows yellow circles in the PR next to in progress checks, whereas there is no yellow circle I'm aware of on travis, just two yellow circles that are horizontally aligned or else spinning around a common center.
a
on the github pr page, next to the most recent commit which should have (and did) trigger a travis ci build
hm, it's not shown even after the build has started
e
ok, yeah, this is an existing feature of the github travis integration, think it goes back to inception.
a
the only reason i mention it is because there's no "waiting on 1 checks" anymore
e
That generally indicates flake in Travis API
a
ok, i'll worry if it's longer than a few days
e
u/h, well, if you're talking about, say https://github.com/pantsbuild/pants/pull/5943 this may be yet another nudge to not make such huge changes, or at least changes with so many commits. It may be that there is some scale issue you're hitting - maybe.
a
yes, that makes it incredibly difficult to review and i am very uncomfortably aware of the size of that pr. it makes it harder for everyone, and it's very possible for me to do better, and i will in the future.
e
At the risk of making further review harder (if folks are still keeping track and just want to review changes since they last reviewed), it would be interesting to see if a 
git rebase
that squashed your branch to 1 commit with a force push would get the travis integration working.
if you keep track of the tip sha prior to rebase, you can always reset to that.
a
anecdotally, my much smaller pr #6022 is also seeing this
so i could try it but not sure it's size related at least for the size of the specific PR
e
Agreed. My https://github.com/pantsbuild/pants/pull/6054 just went stale in the same way, so I think one side or both of the integration is hiccuping right now.
a
ok, cool -- seeing at least one more person with it makes me less concerned
(I set up travis last night on a separate repo which shouldn't have affected this at all but that was my worry)
thanks for investigating!
e
Aha. Yeah I've done that on several of my personals with no ill effects
a
phew
e
While I've got you
a
yes!
e
I assume that if we ran docker images in linux ci, this would have made your very slow debugging cycle on crti.o much quicker since it'd be local.
a
yes, and i have a stash that tries to build the travis ubuntu image, but found it difficult to replicate the error cases. i was trying to find information about the travis environment but found it difficult, but didn't try too hard at that time
e
ok, I'm all in ci fixes lately, so I could push on that
a
hm. also, when i did that, i had a much worse understanding of what the failures i was seeing for 5943 were coming from, so maybe i was doing the wrong things to repro
e
Well, we don't use an image on travis right now (except for a rust shard) so what image were you trying?
a
i made a Dockerfile based off of trusty and attempted to hand-apply the fixes in 
.travis.yml
(e.g. using 
jdk_switcher
)
e
ok - yeah, thats a road to insanity
I'll get an issue filed to use our image on all shards
a
ok, that's great to hear (because that's where i was heading)
the 
travis_ci
Dockerfile?
e
yes, but importantly, a Dockerfile we control
That one may need a new including layer or we may need a whole new one, but we'll own the environment fully except for the running kernel
a
yeah, that would be pretty neat
e
k
a
(also, didn't realize dockerfiles of course don't wrap a kernel)
e
Yeah, the host kernel pokes through.
a
as in, if we were to use a dockerfile, that would make me incredibly happy, to clarify
e
Yup
I set this 1/2 up for pex for similar reasons on my end
a
great
i'm not super worried about osx as i have not had any errors beyond files being in different places, which is solved with 5943 searching around for content in a structured way
so this would be truly great
2 Views
Open in Slack
PreviousNext
Powered by