-
Notifications
You must be signed in to change notification settings - Fork 89
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
Add GitHub Actions CI workflow #309
Conversation
Once everything is working well (including code coverage) for release versions, I'll add another workflow that allows failure for nightly, as well as a badge. |
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.
Awesome, thanks for doing this! Looks great!
Coveralls appears to be poorly supported by Coverage outside of Travis. Can we switch to codecov? |
Sure, i can never remember which is which. |
Thanks for this!!
👍 Before we turn off Travis, we need to figure out how to have GA run CI nightly and notify us of failures i.e. the GA version of the current Travis Cron job and
|
That shouldn't be hard to do. Is the cron job only run on master or also for PRs? Where is the cron job specified?
This is trickier, as GA's notification options are more limited. I'll see if I can figure it out with a 3rd party action. |
Something like this seems to be the recommended approach: https://github.saobby.my.eu.orgmunity/t/no-email-notification-on-schedule-cron-job-failure/119638/2. i.e. create a new e-mail address specifically for e-mailing these notifications, and add its password to secrets. This is really inconvenient. |
Codecov Report
@@ Coverage Diff @@
## master #309 +/- ##
=========================================
Coverage ? 97.15%
=========================================
Files ? 16
Lines ? 880
Branches ? 0
=========================================
Hits ? 855
Misses ? 25
Partials ? 0 Continue to review full report at Codecov.
|
Ooof. Thanks for looking into it! Maybe let's leave it for now? ...I wonder if we can just keeping travis for the nightly CI run only? Don't think we need to solve it in this PR. :) |
Okay, this PR is ready for review. A few notes:
|
@mattBrzezinski can i leave it to you to see this merged? |
The changes make sense, not sure how the owners want to handle the failures of the builds. |
There are two different failures:
julia> Δz = Composite{Tuple{Float64, DoesNotExist}}(1.0, Zero())
Composite{Tuple{Float64, DoesNotExist}}(1.0, Zero())
FiniteDifferences v0.11.3 changed how `to_vec` processes this differential, which causes the failure:
julia> to_vec(Δz)[1] # v0.11.2
2-element Vector{Float64}:
-0.06132606674859018
0.0
julia> to_vec(Δz)[1] # v0.11.3
1-element Vector{Float64}:
-0.06132606674859018 @willtebbutt can you comment on whether this is an FD bug or whether we need to use a different differential for the test here? |
Possibly we need to use a matrix with larger values to avoid issues near zero? |
The reported new behaviour is correct -- the previous behaviour was incorrect ( Since we're still treating |
This PR adds a GitHub Actions for CI. The goal is ultimately to replace Travis, though I recommend keeping both in parallel for a little while until we are sure we are happy with this.
cc @simeonschaub