Something is wrong with the meta travis builds about meta HOT 20 CLOSED

Of course.

El vie, 10 de jul de 2015 12:08, gnzlbg [email protected] escribió:

Seems to be related to the biicode part of the travis.yml file in
after_success. @Manu343726 https://github.com/Manu343726 could you take
a look?

—
Reply to this email directly or view it on GitHub
#26.

from meta.

Seems not to be biicode related. It looks like a weird error when trying to clean build artifacts: A lot of warnings raise during git clean -df command to prepare for bii deploy. I will submit an issue to travis guys asking for feedback.

from meta.

This is how it looked before: https://travis-ci.org/ericniebler/meta/builds/69597163#L570

I changed the way we get libcxx in that last PR to use the same script as range-v3, I'll check if that might be causing the problem.

from meta.

Is that the latest successful build?

from meta.

Yes.

The PR build was fine, see here. In the raw log however, it finishes fine.

In the current build, the raw log says at the end that git clean -df finished, but:

travis_time:end:03004b03:start=1436522107387629122,finish=1436522109607647729,duration=2220018607
�[0Ktravis_fold:end:after_success.3
�[0Ktravis_fold:start:after_success.4
�[0Ktravis_time:start:1ae3d0a4
�[0K$ bii init -l
�[32mSuccessfully initialized biicode project �[0m
�[0mtravis_time:end:1ae3d0a4:start=1436522109614835966,finish=1436522111170966689,duration=1556130723
�[0Ktravis_fold:end:after_success.4
�[0Ktravis_fold:start:after_success.5
�[0Ktravis_time:start:0ccfa6be
�[0K$ bii user $BIICODE_USER -p $BIICODE_PASSWORD
INFO: Change user from None to ericniebler �[0m
�[0mtravis_time:end:0ccfa6be:start=1436522111178814864,finish=1436522112799407783,duration=1620592919
�[0Ktravis_fold:end:after_success.5
�[0Ktravis_fold:start:after_success.6
�[0Ktravis_time:start:25e9b382
�[0K$ if [ -n "$BIICODE_PASSWORD" ]; then bii publish -r; fi
�[1m�[33m�[40mWARN: Discarding 'git://github.com/ericniebler/meta.git' origin. Pending changes not pushed!�[0m
INFO: Input origin info:�[0m
�[32mUrl: �[0m

No output has been received in the last 50m0s, this potentially indicates a stalled build or something wrong with the build itself.

The build has been terminated

This is was I assumed that it was biicode related. The bii publish -r command seems to stall for 50 minutes. Maybe it finishes fine but something else after it stalls, I don't know, it's weird.

from meta.

I think I got it: bii publish -r command calls git to add versioning info to the deploy, ao you can see to what commit corresponds your latest "release" available at bii cloud. Seems like the repo is in an invalid state (Maybe the reason for the warnings during clean) so that internal call to git asking for info fails. When this happens bii asks user to enter git remote info manually. That's why the build hangs (Note the last line from the log, 32m waiting for "Url:" input).

This is the function that requests git info, called from bii publish through this function when -r option is passed. The latest function I linked is which asks for user input if git command fails, where the build is locked.

Hopefully (for me xD) this is not bii related but something weird happening with the git repo state. I will continue looking to find the problem.

from meta.

The last build broke with some git related issues:

https://travis-ci.org/ericniebler/meta/builds/70091012

Maybe that is the reason?

from meta.

The last build broke with some git related issues:
https://travis-ci.org/ericniebler/meta/builds/70091012

I've seen that before. It's intermittent and has never causes lasting trouble.

from meta.

Did you guys rebase the doc/gh-pages branch? This problem usually happens to me when I push a feature branch to Travis for testing, then remove that branch after merging to master, and then try to replay the Travis build.

from meta.

The last build broke with some git related issues:
https://travis-ci.org/ericniebler/meta/builds/70091012

This happened to me yesterday too after updating my fork to work on this. See https://travis-ci.org/Manu343726/meta/jobs/70441741

I'm writing an issue to travis-ci repo.

from meta.

Would merging a squashed PR cause this?

from meta.

Would merging a squashed PR cause this?

I don't think so. The situation I was talking about requires modifying public history, which you didn't do.

from meta.

Maybe we should encourage Travis team to talk here, they dropped some feedback at the issue above.

from meta.

Oh man. This was user error. My bad.

from meta.

Now there is a problem with clang-3.4 on travis. @gnzlbg any thoughts?

from meta.

TODO: figure out what's going on with clang-3.4 on travis.

from meta.

FWIW, you may want to cherry-pick from Hana's .travis.yml. It builds with Clang 3.{5,6,7} with libc++ and libstdc++. And it's on the new container-based Travis architecture. It should be pretty trivial to add Clang 3.4 to it for Meta.

from meta.

Thanks @ldionne , gonna move meta to the new container-based architecture and see if it solves the problem. Clang 3.7 is also failing for no apparent reason (while for range-v3 it works fine).

from meta.

If nobody has time to look at this (not blaming anybody, time is a scarce commodity), I'll disable travis and biicode integration.

from meta.

Disable biicode deploy if you like, just to be sure the issues are not related to bii. Also this would clear a bit travis builds output.

from meta.