[test][tvos] Disable InvalidIVSizes overflow case for tvos

[mdh1418]

This test has flakily caused app crashes on tvOS arm64 lanes. dotnet/arcade#11167
When the app crashes, it was observed that of the last tests ran, there commonly was only 2 InvalidIVSizes see
https://gist.github.com/mdh1418/563ce4066e16dfee055e0903e2c70a1e
https://gist.github.com/mdh1418/ed11c7ba361c3fdb8906e57034c58f90
https://gist.github.com/mdh1418/c831b807dc0d949bc01cdeee61a1d795

Whereas on a successful test suite run, there are 3.
https://gist.github.com/mdh1418/912914871d580475751d460719624224

Its suspected that the last test case is problematic, so disabling to make CI cleaner

[ghost]

Tagging subscribers to 'os-tvos': @steveisok, @akoeplinger
See info in area-owners.md if you want to be subscribed.

Issue Details

---

This test has flakily caused app crashes on tvOS arm64 lanes. dotnet/arcade#11167
When the app crashes, it was observed that of the last tests ran, there commonly was only 2 InvalidIVSizes see
https://gist.github.com/mdh1418/563ce4066e16dfee055e0903e2c70a1e
https://gist.github.com/mdh1418/ed11c7ba361c3fdb8906e57034c58f90
https://gist.github.com/mdh1418/c831b807dc0d949bc01cdeee61a1d795

Whereas on a successful test suite run, there are 3.
https://gist.github.com/mdh1418/912914871d580475751d460719624224

Its suspected that the last test case is problematic, so disabling to make CI cleaner

Author:	mdh1418
Assignees:	-
Labels:	os-tvos
Milestone:	-

[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

---

This test has flakily caused app crashes on tvOS arm64 lanes. dotnet/arcade#11167
When the app crashes, it was observed that of the last tests ran, there commonly was only 2 InvalidIVSizes see
https://gist.github.com/mdh1418/563ce4066e16dfee055e0903e2c70a1e
https://gist.github.com/mdh1418/ed11c7ba361c3fdb8906e57034c58f90
https://gist.github.com/mdh1418/c831b807dc0d949bc01cdeee61a1d795

Whereas on a successful test suite run, there are 3.
https://gist.github.com/mdh1418/912914871d580475751d460719624224

Its suspected that the last test case is problematic, so disabling to make CI cleaner

Author:	mdh1418
Assignees:	-
Labels:	area-System.Security, os-tvos
Milestone:	-

[mdh1418]

/azp run runtime-ioslike

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[mdh1418]

I'm planning to rerun the tvOS arm64 lane multiple times to see whether or not the System.Security.Cryptograph.Tests suite crashes

[steveisok]

@bartonjs this seems to happen somewhat randomly. The logs aren't capturing the crash with any detail. Can you speculate as to what might be happening?

[vcsjones]

If this is indeed the culprit and we need to unblock CI here can we use an ActiveIssue and open an issue instead of a conditional to figure out why it's failing?

I can look at this later.

EDIT: oh it's a theory. I guess you can't put an ActiveIssue on just one case. I still think we should open a tracking issue and put that link in the skip exception.

[mdh1418]

@vcsjones I'm not certain which test is the actual one causing the crash so I'll create an issue and add it after running it a few times to ensure that its no longer crashing if that test case is skipped

Edit: Actually, I'll add the issue and link it in because I forgot to include the XUnitExtensions import anyways.

[mdh1418]

/azp run runtime-ioslike

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[bartonjs]

Can you speculate as to what might be happening?

I don't think there's anything wrong with the crypto code, but rather the runtime or the test harness/runner.

All the test does is:

• Create a gigantic array (536870928 bytes).
  • If that throws OutOfMemoryException, OK.
• Call Aes.Create().
  • Maybe we've pushed the limit here and we're OOMing, but the object isn't very big, so it seems unlikely. In that case I'd expect the test runner to still catch/log the OOM.
• Pass the gigantic array to CreateEncryptor.
  • That clones the array (another 536870928 bytes, making for a total test demand of 1,073,741,856 bytes), then it's shown to be the wrong size and discarded. (literally using Object.Clone(), if that matters)
    • If this threw OOM the test would pass.
    • Then the cloned array goes out of scope and can be GC'd.
• Pass the gigantic array to CreateDecryptor
  • Same as with CreateEncryptor, we get the peak demand of just over 1GB, and an OOM would pass.

So it seems likely that something during the execution of this test is falling over due to a low memory condition.

Yeah, we could change the product code to notice that the array is the wrong size before cloning it, but that doesn't seem to address that we have either a CLR problem where the system doesn't tolerate low memory properly or a test harness problem where we're not capable of capturing this condition/state.

[mdh1418]

/azp run runtime-ioslike

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[mdh1418]

/azp run runtime-ioslike

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[mdh1418]

Of the 3 runs of runtime-ioslike which runs tvOS arm64's System.Security.Cryptography.Tests suite, all have not crashed and the test is successfully ignored.

[mdh1418]

/azp run runtime-extra-platforms

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[akoeplinger]

I think we've seen similar issues in other test cases where gigantic arrays / OOM are involved, it looks like a more general Mono issue on low-memory devices. I'm fine with this as a quick fix but we should probably investigate more about the root cause.

[mdh1418]

The remaining failures on runtime-extra-platforms are
#76830
dotnet/arcade#11210
dotnet/xharness#964
#76501