travis: add Bionic (Ubuntu 18.04) to test matrix#3606
travis: add Bionic (Ubuntu 18.04) to test matrix#3606estesp merged 3 commits intocontainerd:masterfrom
Conversation
|
Okay, not yet there; it's now running the following Linux combinations:
So I either have to write-out all combinations, or at least 3 for xenial |
265816b to
874a4a4
Compare
|
Build succeeded.
|
Codecov Report
@@ Coverage Diff @@
## master #3606 +/- ##
==========================================
+ Coverage 42.38% 42.43% +0.04%
==========================================
Files 126 126
Lines 13900 13919 +19
==========================================
+ Hits 5891 5906 +15
- Misses 7123 7125 +2
- Partials 886 888 +2
Continue to review full report at Codecov.
|
|
All working now:
|
|
The only difficulty is that travis doesn't clearly show what distro each build is in the overview page; I added a dummy |
|
@AkihiroSuda this is working now; ptal 👍 |
|
Build succeeded.
|
24018dc to
5a446ac
Compare
|
force-pushed to trigger CI again (linter took to long) |
|
Build succeeded.
|
|
👍 all green now |
|
This increases test time for a PR from 20min to 40min. I'm not sure its worth the change, we want our PRs to be fast and provide quick feedback to users. Since containerd does not code to a specific distribution, I don't think it makes sense to have a matrix for this. |
|
This runs 7 jobs per PR and merge to master and we get 5 parallel from Travis; similar to ppc64le (which also had a perf. issue) means longer PR runs especially when we get multiple PRs in process. Any chance we could minimize runs on the older distro (less of the runtime shim variants maybe?) |
|
Ah, didn't know there was a limit of 5 parallel. I initially wanted to update to Bionic, but @AkihiroSuda mentioned we may want to also test on the older LTS, that's why I added the matrix; I can remove two of those, or create different combinations |
|
Any preference which runtime(s) to test on Xenial / Bionic? |
|
Can we make it so that master tests on all of them but PRs only run on the new ones? |
|
good one; let me google a bit if travis can do different things based on the branch |
|
looks like there's such an option; travis-ci/travis-ci#7149 (comment) |
Using bionic (current LTS) as default, and add xenial (Ubuntu 16.04 LTS) to the matrix, to test the previous LTS release as well on master Signed-off-by: Sebastiaan van Stijn <[email protected]>
Signed-off-by: Sebastiaan van Stijn <[email protected]>
5a446ac to
94e4b68
Compare
|
Build succeeded.
|
The branch (for pull requests), is the base/target branch, so for pull requests against master, these would still be run. From the travis documentation: > branch (the current branch name; for pull requests: the base branch name) This patch excludes these jobs by not running them for pull request (event type=pull_request or event type=push (when rebasing a pull request)). Signed-off-by: Sebastiaan van Stijn <[email protected]>
|
I think I have it setup correctly now; just filtering on Not sure how to test if it now correctly runs on master (without merging this PR) |
|
Build succeeded.
|
|
@crosbymichael @estesp ptal |
|
LGTM |
|
hmm.. https://travis-ci.org/containerd/containerd/builds/580419692 Maybe needs some more tweaking to get master merge runs on xenial added? |
Ah booh! The docs is quite confusing on this 😞, and I have no idea how to test before actually merging the changes 😢 I wonder what type of event is needed to run in on "master" (would that be a Also, looking at this again; filtering on if: branch = master AND type NOT IN (push, pull_request)Perhaps just this would do; if: type NOT IN (pull_request) |
|
I think it's a push event on master |
|
yup; changed it in #3614 and looks like that's working |
Using bionic (current LTS) as default, and adding
xenial(Ubuntu 16.04 LTS) to the matrix, to test the previous LTS release as well(see discussion on rootless-containers/usernetes#111)