[stdlib] Remove #warning in Random.cpp (_WIN32)

[benrimmington]

No description provided.

[benrimmington]

@compnerd Have you managed to test swift_stdlib_random on Windows?

Please merge this pull request, if you think the #warning isn't needed.

[compnerd]

@benrimmington no, I don't think that we have tests for this yet which is why I haven't removed the warning yet.

[benrimmington]

swift_stdlib_random is used in SystemRandomNumberGenerator._fill(bytes:), which has a validation-test.

Should the _fill(bytes:) customization point be removed in Swift 5? It doesn't appear to be used elsewhere (since #17604).

Could the validation-test call swift_stdlib_random (exported via SWIFT_RUNTIME_STDLIB_API) directly?

Cc: @mikeash, @Azoy

[Azoy]

@benrimmington I have a ready implementation to formally introduce fillBytes(_:), but I haven't gotten around to writing the actual proposal. I'm not sure even if I had something ready by tomorrow that it'd go through review before Nov. 16 however. I've been very busy 😞

[benrimmington]

@Azoy Can your fillBytes(_:) proposal be reviewed for Swift 5.1 or later? It might be able to use the new DataProtocol.

Does your next(upperBound:) amendment need to be reviewed by next week?

Shall I remove _fill(bytes:) now, so that it isn't in the Swift 5.0 ABI?

Shall I also remove the _fill(bytes:) validation-test? Other tests are indirectly using swift_stdlib_random.

[mikeash]

If it's not used and has an underscore prefix then it seems like _fill(bytes:) is just an attractive nuisance and it would be best to remove it.

[Azoy]

cc: @lorentey who might have some opinions on what we could do.

[Azoy]

cc: @airspeedswift as well

[lorentey]

Oh, I haven't realized we aren't using _fill(bytes:) anymore. Removing it seems like a good idea indeed!

As long as we don't need it, It seems cleaner to add a de-underscored version later than trying to support both.

[lorentey]

I submitted #20463 to remove it!