-
Notifications
You must be signed in to change notification settings - Fork 3.8k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[ci] changed CRAN mirror to fix failing installs #2954
Conversation
ugh same error on Travis as #2949 (comment). When it worked on my Mac I was so ready to believe it was transient 😫 But the Azure DevOps build succeeded 0_o |
I am running the same version of R on my Mac and you can see from my screenshot that it clearly worked there. If you click around in a browser at https://cran.rstudio.com/bin/macosx/el-capitan/contrib/ you can see that all the folders are now empty, but all the packages we need are there at https://cloud.r-project.org/bin/macosx/el-capitan/contrib/3.6/. I am confused |
@jameslamb |
oh great! it seems like this PR is passing now. Did you restart the |
Yep :-) Let me rerun it several times to confirm we are not facing it anymore. |
I re-run 3 times - the third one is errored! |
Exactly the same versions of installed packages in green and red builds... |
ah!!! I'm afraid 😬 I traced a chain of comments on errors from conda/conda#8838 to conda-forge/libarchive-feedstock#35 (comment) where I see
I can see in this failed build we ended up with
In this successful build we got the same thing:
WHAT IS HAPPENING haha |
@StrikerRUS if you put this back (22d4076) I wonder if it would work? Even though it's sometimes redundant, it would guarantee we have |
I'll create a new branch and try several rebuilds there to not pollute other PRs. I'll get back to you ASAP (~3-5 re-runs of |
Unfortunately, no difference with excplicitly installed |
I don't see
^ nevermind, I can do this on my fork |
Just re-login: #2827 (comment).
Hmm, there are so many issues with CIs in these days, I'm not sure in anything 🙁 For example, right now Travis status is yellow while all jobs have finished already. I'll try to do it again. |
I added an explicit install of
I'm going to try to get to 10 consecutive builds, and if they work I'll add that to this PR |
At this point Travis stopped loading for me. I'm worried that the service is down :/. It's had some struggles the past week https://www.traviscistatus.com/ That would explain why we're seeing this: They had an incident related to it yesterday |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@jameslamb OK, let's take a break and get back to CI errors tomorrow.
I'm approving this PR. Feel free to add explicit conda installation to the lint
job if think it can help and merge the PR (if Travis will let you do that 😄 ).
🤝 deal, I'll push that change and then let's come back to it tomorrow. Thanks for the help! |
Travis definitely did have an outage today! (link) This went up a bit after we stopped working on this: I'll rebuild two more times on my personal to make sure this is change works, then merge. |
ok now that the Travis outage seems to be behind us, did 3 more consecutive builds, all passed the
I feel ok with merging this. |
This pull request has been automatically locked since there has not been any recent activity since it was closed. To start a new related discussion, open a new issue at https://github.com/microsoft/LightGBM/issues including a reference to this. |
This fixes the issue we are seeing in Mac builds across all PRs right now.
@StrikerRUS I know you tried a similar change in #2949 (comment) and said it didn't work. I don't see anything obviously wrong with your approach, but I do know the change I'm suggesting in this PR works on my Mac.
Changes:
.Rprofile
Since we only have one
install.packages()
call, I don't think we need to deal with the complexity of when and if the.Rprofile
file gets sourced.on my Mac:
My best guess about what happened is:
Even if that is true and a few hours from now the RStudio repo will be back up, I still think this PR should be accepted. It removes unnecessary indirection (with
.Rprofile
) and points our builds at the real source of truth for R packages.