Disabled System.Security.Cryptography.Tests for Android x64 and x86

[ilonatommy]

Fixes #66831.

[ghost]

Tagging subscribers to this area: @dotnet/area-system-security, @vcsjones
See info in area-owners.md if you want to be subscribed.

Issue Details

---

Fixes #66831.

Author:	ilonatommy
Assignees:	ilonatommy
Labels:	area-System.Security
Milestone:	-

[mdh1418]

Thanks, looks good to me after fixing the paths and whether we want it to still run on Android Arm64.

[bartonjs]

Disabling crypto tests on Android feels like a very heavy hammer. Especially since this is one of the areas where there's actually a lot of OS-specific code (so it's significantly more important to run the crypto tests on Android than it is to run, e.g. collections tests).

Are we running out of memory because there's a memory leak? If so, someone (with os-android experience) should investigate/fix it.

Is it just intermittent because it's down to test parallelization? Maybe disable test parallelization for the library (on Android).

[ilonatommy]

For now I disabled only the ones that actually posed a problem, not all of them, taking into consideration @bartonjs's opinion. That's only a temporary solution till the fix won't be found. @bartonjs, @danmoseley, @mdh1418 do you know who might be able to deal with this issue?

[mdh1418]

I'm not sure who is the go to person for a memory issue like this, but if we follow through with just disabling the problematic architectures and keeping the Arm64 in hopes that it would catch all possible regressions, the changes look good to me.

[danmoseley]

On our team we do not have equipment or expertise to investigate device specific issues. Mono team handles this. @mdh1418 who on your team would look?

[steveisok]

Disabling crypto tests on Android feels like a very heavy hammer. Especially since this is one of the areas where there's actually a lot of OS-specific code (so it's significantly more important to run the crypto tests on Android than it is to run, e.g. collections tests).

Are we running out of memory because there's a memory leak? If so, someone (with os-android experience) should investigate/fix it.

Is it just intermittent because it's down to test parallelization? Maybe disable test parallelization for the library (on Android).

I agree it's too heavy a hammer. I don't believe it's a memory leak. We've never been able to reproduce this outside of CI.

I think disabling test parallelization for this suite is something we should try. Disabling test randomization would be another thing to add as it may highlight where it fails more often.

[steveisok]

On our team we do not have equipment or expertise to investigate device specific issues. Mono team handles this. @mdh1418 who on your team would look?

If we come up empty disabling parallel test execution, my team can take a closer look. Since we've never been able to reproduce locally, I suspect it'll be easier to look at the tests first and see if there's anything we can do to reduce contention. We would need @bartonjs to help us if we get to that point.