Added arm64 jobs for Travis-CI #18000
Conversation
|
As suggested in #16879, tried removing pandas but it was giving test errors. So, installing pandas using conda to resolve the timeout error and added some dependencies as well. Please review and let me know if any changes required. |
| @@ -184,10 +216,10 @@ script: | |||
| # Each script we want to run need to go in its own section and the program | |||
| # you want to fail travis needs to be the last thing called. | |||
| - | | |||
| # The number of processes is hardcoded (-n2), because using too many | |||
| # The number of processes is hardcoded (-n4), because using too many | |||
There was a problem hiding this comment.
we get 4 usable cores on travis now?
There was a problem hiding this comment.
In Arm64, we can use more than two cores, please refer "https://travis-ci.community/t/nproc-reports-32-cores-on-arm64/5851/2". In amd64 it is picking 4 workers as well, but it maybe running on 2 cores only.
but it("-n4") has significant difference in time for Arm64 platform.
There was a problem hiding this comment.
The core count wasn't the problem, it was opening too many Inkscape and Ghostscript processes, as noted in the comment below. Though now we pipe everything into a single Inkscape instance (and Ghostscript?) per process, so maybe that's not an issue.
.travis.yml
Outdated
| @@ -131,6 +151,18 @@ install: | |||
| git describe | |||
| # Upgrade pip and setuptools and wheel to get as clean an install as possible. | |||
| python -mpip install --upgrade pip setuptools wheel | |||
| - | | |||
| # Using conda to install pandas, pyzmq & ipython | |||
There was a problem hiding this comment.
can we skip the tests the rely on pandas and jupyter instead? I would rather not bring conda / conda-forge into the mix on travis if possible...
There was a problem hiding this comment.
Sure, we can skip those tests for Arm64 platform. Updated the same.
d4d1c90 to
40f51f0
Compare
b4d288b to
e7651dd
Compare
| @@ -184,10 +216,10 @@ script: | |||
| # Each script we want to run need to go in its own section and the program | |||
| # you want to fail travis needs to be the last thing called. | |||
| - | | |||
| # The number of processes is hardcoded (-n2), because using too many | |||
| # The number of processes is hardcoded (-n4), because using too many | |||
There was a problem hiding this comment.
The core count wasn't the problem, it was opening too many Inkscape and Ghostscript processes, as noted in the comment below. Though now we pipe everything into a single Inkscape instance (and Ghostscript?) per process, so maybe that's not an issue.
11bdcff to
54333d9
Compare
| @@ -154,6 +154,10 @@ def check_alt_backend(alt_backend): | |||
| @pytest.mark.parametrize("toolbar", ["toolbar2", "toolmanager"]) | |||
| @pytest.mark.flaky(reruns=3) | |||
| def test_interactive_backend(backend, toolbar): | |||
| # Skipping the test for Travis-CI on arm64 platform | |||
| if os.environ.get('TRAVIS_CPU_ARCH') == "arm64": | |||
There was a problem hiding this comment.
Does this need a default value (ie. os.environ.get('TRAVIS_CPU_ARCH', '') in case this environment variable isn't present?
b731d7d to
f058b6a
Compare
Signed-off-by: odidev <odidev@puresoftware.com>
f058b6a to
6ac1e6e
Compare
|
Thanks for working on this. Unfortunately, Travis seems to be kind of dead for us, so I'm going to close this PR. We do have wheels for aarch64 added in #19402. |
Uh oh!
There was an error while loading. Please reload this page.