Bug 239057 – [WebGPU] WebGPU strings are UTF-8, not Latin-1

• Status: RESOLVED
• Resolution: FIXED
• Component: WebGPU

[Myles C. Maxfield]

.

[Myles C. Maxfield]

Created attachment 457219 [details]
Patch

[Kimmo Kinnunen]

Comment on attachment 457219 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=457219&action=review

> Source/WebGPU/WebGPU/Adapter.mm:137
> +    auto label = String::fromUTF8(descriptor.label);

Aren't these coming from the API? E.g. would fromAPI() be more appropriate?

[Myles C. Maxfield]

Comment on attachment 457219 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=457219&action=review

>> Source/WebGPU/WebGPU/Adapter.mm:137
>> +    auto label = String::fromUTF8(descriptor.label);
> 
> Aren't these coming from the API? E.g. would fromAPI() be more appropriate?

Good point.

[Myles C. Maxfield]

Committed r292748 (?): <https://commits.webkit.org/r292748>

[Radar WebKit Bug Importer]

<rdar://problem/91598857>

[Darin Adler]

Comment on attachment 457219 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=457219&action=review

> Source/WebGPU/ChangeLog:8
> +        Replace String::fromLatin1() with String::fromUTF8().

Great to fix this if it was going to mishandle non-ASCII characters. But note that now there is a failure case for all of these, if it’s invalid UTF-8. This is a major difference between String::fromUTF8 and String::fromLatin1, and might eventually lead to us renaming it to String::tryFromUTF8.

As long as each caller either has a guarantee it will be valid UTF-8, or can withstand a null string result when it’s not, we are OK.

[Myles C. Maxfield]

Comment on attachment 457219 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=457219&action=review

>> Source/WebGPU/ChangeLog:8
>> +        Replace String::fromLatin1() with String::fromUTF8().
> 
> Great to fix this if it was going to mishandle non-ASCII characters. But note that now there is a failure case for all of these, if it’s invalid UTF-8. This is a major difference between String::fromUTF8 and String::fromLatin1, and might eventually lead to us renaming it to String::tryFromUTF8.
> 
> As long as each caller either has a guarantee it will be valid UTF-8, or can withstand a null string result when it’s not, we are OK.

Right. For this situation, it's the latter.