You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Unless I'm missing something the aarch64 TravisCI job is actually running a x86_64/amd64 VM, not an ARM one.
This can be seen for instance by the wheels names that are downloaded to run tests in CI,
Using cached mypy-0.782-cp37-cp37m-manylinux1_x86_64.whl (20.8 MB)
(unlike for e.g. s390x or ppc64le jobs).
The fix would be to specify "arm64" instead of "aarch64" in .travis.yml.
This is due to an unfortunate approach by TravisCI to fall back to amd64 for unsupported architectures, instead of failing (as documented in their docs). The "aarch64" shown in the UI does not seem reliable either, but one can use $TRAVIS_CPU_ARCH environment variable to double check.
Maybe they renamed this arch at some point, not sure.
Indeed. You can see in a build log from a recent PR (expand the "Build system information" to see line 39) that the machine architecture is linux/amd64, even though the architecture name is aarch64. Thanks!
Unless I'm missing something the aarch64 TravisCI job is actually running a x86_64/amd64 VM, not an ARM one.
This can be seen for instance by the wheels names that are downloaded to run tests in CI,
(unlike for e.g. s390x or ppc64le jobs).
The fix would be to specify "arm64" instead of "aarch64" in
.travis.yml
.This is due to an unfortunate approach by TravisCI to fall back to amd64 for unsupported architectures, instead of failing (as documented in their docs). The "aarch64" shown in the UI does not seem reliable either, but one can use
$TRAVIS_CPU_ARCH
environment variable to double check.Maybe they renamed this arch at some point, not sure.
Discovered while trying to add a similar setup in scikit-learn/scikit-learn#17996
cc @mattip
The text was updated successfully, but these errors were encountered: