Fabian Meumertzheim
Fabian Meumertzheim
If your use case really doesn't need to specify `exec_compatible_with` on a per-target level, you could register a toolchain with the required (global) constraint and a bit of undocumented logic...
> FWIW I tried implementing a @bazel_tools//tools/cpp:test_runner_toolchain_type and it seems to suffer from the same issue that it can't affect only the test action's exec platform. It should be able...
@Wyverald Could you help cherry-pick this? It looks like there are a number of refactoring commits missing (AutoValue -> record, module thread context extraction, ...) but I don't know which...
This is motivated by https://github.com/bazelbuild/bazel/pull/23328, in which a number of tests based on `genrule` I added failed open due to an experimental flag not propagating to the exec configuration.
@lberki Could you review this as Bazel's "Chief Flag Officer"? :-) @aranguyen for configurability
@lberki Friendly ping
I added a test for incompatible tests about a year ago, but back then I wasn't sure whether experimental flags should be treated in the same way.
> Do we need re-review? Sorry, clicked the button accidentally while going over my pending PRs.
@lberki Is this feature supported by the internal test strategy implementation? Could you also let me know what `test.warnings` is about?