[iOS/tvOS] RegexCultureTests.Match_In_Different_Cultures test failures

[steveisok]

After #60607 was merged, certain theory runs started failing in the Match_In_Different_Cultures test. Unsure as to why at this point, but here is an example failure:

<test name="System.Text.RegularExpressions.Tests.RegexCultureTests.Match_In_Different_Cultures(pattern: \&quot;(?i:iI+)\&quot;, options: None, culture: , input: \&quot;abcIiİxyz\&quot;, match_expected: \&quot;Iiİ\&quot;)" type="System.Text.RegularExpressions.Tests.RegexCultureTests" method="Match_In_Different_Cultures" time="0.0057729" result="Fail">
        <failure exception-type="Xunit.Sdk.EqualException">
          <message><![CDATA[Assert.Equal() Failure\n            ↓ (pos 2)\nExpected: Iiİ\nActual:   Ii\n            ↑ (pos 2)]]></message>
          <stack-trace><![CDATA[   at System.Text.RegularExpressions.Tests.RegexCultureTests.Match_In_Different_Cultures(String pattern, RegexOptions options, CultureInfo culture, String input, String match_expected)
   at System.Reflection.RuntimeMethodInfo.InvokeWorker(Object obj, BindingFlags invokeAttr, Span`1& parameters)]]></stack-trace>
        </failure>
      </test>

The rest can be found in https://gist.github.com/steveisok/702d94768c660462f0df76889c877e71

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

Issue Details

---

After #60607 was merged, certain theory runs started failing in the Match_In_Different_Cultures test. Unsure as to why at this point, but here is an example failure:

<test name="System.Text.RegularExpressions.Tests.RegexCultureTests.Match_In_Different_Cultures(pattern: \&quot;(?i:iI+)\&quot;, options: None, culture: , input: \&quot;abcIiİxyz\&quot;, match_expected: \&quot;Iiİ\&quot;)" type="System.Text.RegularExpressions.Tests.RegexCultureTests" method="Match_In_Different_Cultures" time="0.0057729" result="Fail">
        <failure exception-type="Xunit.Sdk.EqualException">
          <message><![CDATA[Assert.Equal() Failure\n            ↓ (pos 2)\nExpected: Iiİ\nActual:   Ii\n            ↑ (pos 2)]]></message>
          <stack-trace><![CDATA[   at System.Text.RegularExpressions.Tests.RegexCultureTests.Match_In_Different_Cultures(String pattern, RegexOptions options, CultureInfo culture, String input, String match_expected)
   at System.Reflection.RuntimeMethodInfo.InvokeWorker(Object obj, BindingFlags invokeAttr, Span`1& parameters)]]></stack-trace>
        </failure>
      </test>

The rest can be found in https://gist.github.com/steveisok/702d94768c660462f0df76889c877e71

Author:	steveisok
Assignees:	-
Labels:	area-System.Text.RegularExpressions, untriaged
Milestone:	-

[steveisok]

/cc @stephentoub

[veanes]

I just took a look in Match_In_Different_Cultures_TestData(). Original intent with this test (I wrote the original version of this test) was to make sure correct behavior in NonBacktracking. It seems some of those tests have been also lifted to Compiled and None, where the problem is caused by incorrect handling of the Turkish i's: \u0130 and \u0131 that is still an open issue I believe.

[stephentoub]

They were extended here to None and Compiled:
https://github.com/dotnet/runtimelab/pull/1547/files#diff-a3647ce913fdcfbfdfec36e888f7313551577a5b6f1ecf313da00d9471336003R218
But the problem in this issue is that the test is passing on Windows, Linux, and macOS... it's just failing for some inputs on iOS and Android.

[veanes]

Oh, I see. So the test itself is in some sense correct here actually.

[steveisok]

I don't doubt the test itself is correct.

If these failures were just isolated to Android, I wouldn't have even opened an issue as a great deal of tests with tr-TR culture produce different results. iOS has not had that problem, which leads me to believe there may be some bug lurking.

[stephentoub]

@steveisok, do you want us to disable the test for now for Android/iOS?

[steveisok]

That's ok. I have a PR pending that will skip it.

[stephentoub]

@joperezr, was this addressed by your IgnoreCase change?

[joperezr]

good question, not clear. I haven't ran the iOS/tvOS tests locally before but I'll check by re-adding the test and running against Helix.