[BigInteger] Parse test with trailing spaces fails on some cultures

[Maxwe11]

System.Numerics.Tests.parseTest.RunParseToStringTests [FAIL]
    Assert.Throws() Failure
    Expected: typeof(System.FormatException)
    Actual:   (No exception was thrown)
    Stack Trace:
       m:\Workspace\github\corefx\src\System.Runtime.Numerics\tests\BigInteger\parse.cs(408,0): at System.Numerics.Tests.parseTest.VerifyParseToString(String num1, NumberStyles ns, Boolean failureNotExpected, String expected)
       m:\Workspace\github\corefx\src\System.Runtime.Numerics\tests\BigInteger\parse.cs(380,0): at System.Numerics.Tests.parseTest.VerifyParseToString(String num1, NumberStyles ns, Boolean failureNotExpected)
       m:\Workspace\github\corefx\src\System.Runtime.Numerics\tests\BigInteger\parse.cs(247,0): at System.Numerics.Tests.parseTest.VerifyNumberStyles(NumberStyles ns, Random random)
       m:\Workspace\github\corefx\src\System.Runtime.Numerics\tests\BigInteger\parse.cs(33,0): at System.Numerics.Tests.parseTest.RunParseToStringTests()
Finished:    System.Runtime.Numerics.Tests
=== TEST EXECUTION SUMMARY ===
System.Runtime.Numerics.Tests  Total: 2, Errors: 0, Failed: 1, Skipped: 0, Time: 20,751s

Repro steps:

1. Start->Run (WinKey + R) -> intl.cpl
2. Change format (no need to change locale) to Ukrainian (Ukraine)
3. build /p:WithCategories=OuterLoop

Possible solution: dotnet/corefx#1592 (comment)

[AlexGhiondea]

Looks like we know what needs to happen here. It is a simple fix that would be a great place to start looking at the CoreFx code.

[ghost]

I'm looking for an opportunity to get started contributing to .NET Core. Would it be possible to get this issue assigned to me?

[mellinoe]

I'm looking for an opportunity to get started contributing to .NET Core. Would it be possible to get this issue assigned to me?

Sure!

I'll try to give you an updated set of instructions to use to repro this. Since this was filed so long ago (almost 2 years), we're not even sure if it still repros. Chances are it still does, though, because none of us run the tests in a Ukrainian environment as far as I know.

Steps:

• Clone repo
• Run build.cmd
  Run build-tests.cmd -- /p:Outerloop=true

If you then want to repro the individual test project that is failing, you can run

msbuild src\System.Runtime.Numerics\tests\System.Runtime.Numerics.Tests.csproj /t:RebuildAndTest /p:Outerloop=true

[mellinoe]

@karelz I think we need to add @dennisdietrich to the "Contributors" group before I can assign him any issues. How do we do that, again?

[karelz]

Invitation sent (you need super-powers in dotnet org to make it happen).
@dennisdietrich please ping me here, when you accept and I will assign it to you.

[ghost]

Thanks @karelz! Done.

[ghost]

@mellinoe, quick question (I'm sure this is documented somewhere but I can't find it): What about line endings? Should they be Windows or UNIX-style/what value is preferred for core.autocrlf?

[mellinoe]

what value is preferred for core.autocrlf

True, I believe. TBH I've never actively thought of this; we use the default as far as I know.

[ghost]

@mellinoe, the good news (I guess) is that I can repro the issue. Unfortunately, so far I'm unable to build System.Runtime.Numerics.sln in VS (trying to get help on Gitter).

[ghost]

Still stuck trying to build/debug corefx in VS so it's difficult to do a root cause analysis. However, since I can reproduce on the command line I've looked at the stack trace and so far this looks like a legitimate product bug to me (though it is labeled 'test bug'). First of all, the test passes for invariant, en-US, de-DE, and hi-IN. Second, it chokes on trailing 0x20 characters for uk-UA which is just a regular space.

Edit: Checked CultureInfo.NumberFormat for uk-UA but that just seems to confirm my result so far. It is using whitespace characters in a few places but those are 0xA0 and not 0x20.

[ghost]

N/m. Found this method in FormatProvider.Number so I guess this is by-design and it's interpreting the trailing 0x20 as a trailing group separator. Preparing a PR on the basis of this assumption.

private static unsafe char* MatchChars(char* p, char* str)
{
    Debug.Assert(p != null && str != null);

    if (*str == '\0')
    {
        return null;
    }

    // We only hurt the failure case
    // This fix is for French or Kazakh cultures. Since a user cannot type 0xA0 as a
    // space character we use 0x20 space character instead to mean the same.
    while (*p == *str || (*str == '\u00a0' && *p == '\u0020'))
    {
        p++;
        str++;
        if (*str == '\0') return p;
    }
    return null;
}

[karelz]

Still stuck trying to build/debug corefx in VS

Can you provide some details? Maybe @weshaggard @mellinoe can help ...

[ghost]

Already discussing this with Eric on Gitter (seemed like the better place for figuring this out). Thanks!