Use FileId::MAX for id assertion in PathInterner::intern

[Shourya742]

Update PathInterner::intern to use FileId::MAX for id range checks, ensuring FileId MAX invariant.

[Shourya742]

I am curious why we don't use u32::MAX?

[Shourya742]

cc: @ShoyuVanilla apologies to tag, just looking for a quick review on this.

[lnicola]

I am curious why we don't use u32::MAX?

The file id bits are OR-ed with 8 edition bits and a reserved one and get passed to proc macros as a single EditionedFileId.

What I'm wondering is why EditionedFileId::MAX_FILE_ID is 0x007F_FFFF while FileId::MAX is 0x7FFF_FFFF.

[Shourya742]

I am curious why we don't use u32::MAX?

The file id bits are OR-ed with 8 edition bits and a reserved one and get passed to proc macros as a single EditionedFileId.

What I'm wondering is why EditionedFileId::MAX_FILE_ID is 0x007F_FFFF while FileId::MAX is 0x7FFF_FFFF.

It’s interesting that FileId reserves one bit but doesn’t include any edition bits. Since FileId is expected to accommodate editions as well, the mask 0x7FFF_FFFF appears incomplete.

[Shourya742]

@lnicola makes sense to update the FileId upper limit to 0x007F_FFFF, right? What do you think?

[Veykril]

This is definitely more correct, not that we will ever observe issues from this given that r-a likely fails with that many files in memory anyways though. Thanks for finding this!