fix(ci): Re-enable velox_table_evolution_fuzzer_test in CPU builds

[simoneves]

Once facebookincubator/velox#17106 lands, this test can once more be safely run in CI.

[mattgara]

LGTM, is there a passing example run with this enabled?

[simoneves]

We can test this part officially once the fix lands.

[Avinash-Raj]

But this script test_velox.sh is no longer being called from a velox-test.yml CI workflow.

[simoneves]

But this script test_velox.sh is no longer being called from a velox-test.yml CI workflow.

Oops! OK... what runs tests now then? Does it need an equivalent change. Would look but on my phone and too lazy to fire up the computer...

[bdice]

The CPU test command is here:

velox-testing/.github/workflows/velox-test.yml

Lines 87 to 88 in 0f2e363

	docker run --rm "${IMAGE}" bash -c \
	"cd /opt/velox-build/release && ctest -j 16 --label-exclude cuda_driver --output-on-failure --no-tests=error"

If we are supposed to use a common/shared test script, that's fine with me as long as we can call it through Docker with a custom image like the current GitHub Actions setup.

[simoneves]

If we are supposed to use a common/shared test script, that's fine with me as long as we can call it through Docker with a custom image like the current GitHub Actions setup.

If the new YAML implementation does the job, I don't see a need to change it. Presumably this is how we were reminded that the table_evolution_fuzzer test was problematic, as the old script didn't run it. Also presumably there are no issues with the other three tests the old script didn't run either.

If that's now the definitive method, should we remove the test_velox.sh script completely, or leave it for posterity?