{CI testing} #342
{CI testing} #342
Conversation
|
Looks I got the CI fixed to only test the last commit, and not all commits. |
|
test this please |
1 similar comment
|
test this please |
|
Testing with cvxopt and mosek in virtualenv's is now working. |
|
another intentional master push: b1943ee |
|
test this please |
1 similar comment
|
test this please |
|
@whoburg How do I re-open this? |
|
It is so! |
|
Thank you! |
|
Sure thing! let me know when to close it :p |
|
test this please |
|
@whoburg @bqpd As best as I can tell there is definitely a bug inn jenkins. I submitted a bug report here:https://issues.jenkins-ci.org/browse/JENKINS-29963?jql=project%20%3D%20JENKINS%20AND%20status%20in%20(Open%2C%20"In%20Progress"%2C%20Reopened)%20AND%20component%20%3D%20%27ghprb-plugin%27 The temporary solution is to make sure you build all pull requests twice. The testing will also be 5 times faster once we find a way to not use .gpkit, at least in the testing. |
|
Hmm, okay, thanks! |
|
test this please |
|
Ok, I think I was able to fix it so you do not have to build twice. The way the changes were getting computed was what was messing things up. So there is still a bug, but we do not need to build twice to avoid it. |
|
I forgot I need to push something to do a real test. I may have spoken too soon. |
|
I spoke too soon. Keep testing twice. |
|
Haha, will do. Unrelatedly, it'd be great to set up some python3 test...platforms?...at some point. |
|
Yes we can setup python3 testing, and pypy if you want. But until we fix #315 that will increase the testing time by a factor of 4. |
|
test this please |
|
I setup a python3 environment on the Linux machine, and found that there are a lot of failures in the master branch. Let me know when you are ready to have python3 testing. |
|
Will do! |
|
which plugin is the bug likely in? Should we contact the authors? |
|
I have submitted the bug already, but not gotten any response. https://issues.jenkins-ci.org/browse/JENKINS-29963?jql=project%20%3D%20JENKINS%20AND%20status%20in%20(Open%2C%20 This is all people doing things on their own time. Bribery might work, and The plugin repo is here: https://github.com/janinko/ghprb On Thu, 03 Sep 2015, 09:36 Warren Hoburg notifications@github.com wrote:
|
|
test gpkit please |
1 similar comment
|
test gpkit please |
|
test models please |
|
test gpkit please |
2 similar comments
|
test gpkit please |
|
test gpkit please |
|
test this please |
1 similar comment
|
test this please |
|
test models please |
|
current error on windows10 cvxopt is documented here: cvxopt/cvxopt#127 (comment) not sure what the right thing to do here is, since cvxopt just doesn't work with a pip install on the system Python right now. |
|
@bpqd I installed Python 3.7 specifically due to that issue so you could test with that |
|
Wait this is different. So cvxopt only works with Anaconda? Am I reading this correctly? |
|
or if you manually add the MKL libs dir to the dll search path on the system python 😬 |
|
Where do I get the MKL libraries? From intel? |
|
What about the proposed workaround for adding Library/bin to the PATH? |
|
that would work for us, but not for our users, which makes me wonder if we should back to recommending Anaconda to Windows users, in which case we should use Anaconda on the test machine... |
|
Ok I'll work on installing Anaconda some time next week. |
|
I installed anaconda in C:\Anaconda3. Now it looks like there are legitimate unit test errors. |
|
hmm it seems to be saying it can't run the python executable at all?
https://acdl.mit.edu/csi/job/CE_gpkit_PR_unit_tests/3036/buildnode=windows10x64,optimizer=cvxopt/console
…On Mon, Jul 6, 2020 at 9:05 AM Marshall Galbraith ***@***.***> wrote:
I installed anaconda in C:\Anaconda3. Now it looks like there are
legitimate unit test errors.
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
<#342 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AALKAGD5243GHBBZKEGHL4DR2HY45ANCNFSM4BNOW33Q>
.
|
|
You have to change the PATH at the top of the script to: |
|
oh weird it now has a pint error...but only when doing coverage testing. Something about the hashing of our pint cache, potentially? Cannot reproduce, will see if I can with a higher python version |
|
What output are you looking at? |
|
https://acdl.mit.edu/csi/view/convex%20engineering/job/CE_gpkit_Push_unit_tests/735/buildnode=windows10x64,optimizer=cvxopt/console (and if you go back to earlier tests, you'll see it passes when I don't run coverage - only to fail bc coverage is not detected) |
|
Beats me... All I can say is that coverage information is useful and you should try to get this to work. |
|
yeah and I had a user message me this error a month ago but couldn't
reproduce (and they didn't respond), so it's occurring in the wild as well.
… |
|
culprit was pint 0.14 compatibility issues; since that comes installed by default with anaconda it can't be uninstalled. not sure why it was only used by coverage, but seems fixed by 1571e26 |
|
So why are there still failing tests? |
|
the model tests are having a separate issue, but the units tests work now
… |
|
Ok that's good. At least that got resolved. |
I created this pull request so I can mess with getting the CI setup without using other pull requests