Bug 299689 – Clang 21 complains about implicit casting between char32_t and char16_t throughout the codebase

• Status: NEW
• Component: New Bugs

[Ross Kirsling]

As of 271373@main, we decided to turn UChar32 (i.e. int32_t) into char32_t.
More recently, we've also replaced UChar with char16_t (though that part is not an issue).

Clang 21 introduces `character-conversion` warnings for implicit conversion between char32_t and char16_t
(in both directions, citing it "may change the meaning of the represented code unit").

Addressing this will be a similarly-sized endeavor to 271373@main itself, and unfortunately involves patching gtest too.
It took a couple of hours just to get through WTF, PAL, and gtest—see attached patch.

I will temporarily set `-Wno-character-conversion` for PS port just because it's blocking our deployment of extra EWS bots.

[Ross Kirsling]

I'm getting 403 Forbidden when trying to attach a patch; you can find it on Slack instead:
https://webkit.slack.com/archives/CU64U6FDW/p1759041829898719?thread_ts=1758976128.172549&cid=CU64U6FDW

[Ross Kirsling]

`-Wno-character-conversion` has now been temporarily set for PS port:
https://commits.webkit.org/300663@main

[Darin Adler]

Ideally we would come up with a good way to set the warning flags based on clang version rather than platform.

Dave, this new warning is coming up soon in clang 21.

[Michael Catanzaro]

*** Bug 299535 has been marked as a duplicate of this bug. ***

[Michael Catanzaro]

Pull request: https://github.com/WebKit/WebKit/pull/51527

[Michael Catanzaro]

(In reply to Darin Adler from comment #3)
> Ideally we would come up with a good way to set the warning flags based on
> clang version rather than platform.

WEBKIT_PREPEND_GLOBAL_COMPILER_FLAGS checks whether the flag is supported or not, so it's safe to use it unconditionally. Ross proposed this in https://github.com/WebKit/WebKit/pull/51490.

[Darin Adler]

(In reply to Michael Catanzaro from comment #6)
> WEBKIT_PREPEND_GLOBAL_COMPILER_FLAGS checks whether the flag is supported or
> not, so it's safe to use it unconditionally. Ross proposed this in
> https://github.com/WebKit/WebKit/pull/51490.

That’s neat.

Maybe we should turn off this warning immediately using WEBKIT_PREPEND_GLOBAL_COMPILER_FLAGS and then turn the warning back on while patching all the call sites to adapt to it. No reason we have to choose, we can do both!

[Radar WebKit Bug Importer]

<rdar://problem/161980988>