collate: Add support for equivalence classes and collating symbols

[bnovkov]

Summary

This change adds support for matching single and multi-character
equivalence classes and collating symbols, as specified
by POSIX1, section 9.3.5.

Sponsored by: Klara, Inc.
Obtained from: https://github.com/apple-oss-distributions/Libc

[markj]

lib/libc/locale/collate.c

779

What exactly is the intent behind this casting? I guess we have to handle the case where l > slen, but slen is unsigned, so this looks strange.

983

985

lib/libc/locale/collate.h

139

What's the intended consumer of these symbols?

lib/libc/locale/Symbol.map needs to be updated for these symbols to be visible.

[markj]

lib/libc/locale/collate.h

139

Sorry, ignore me, I see that the next revision adds some uses of these within libc.

[bnovkov]

Address @markj 's comments.

lib/libc/locale/collate.c

779

Hm, this does look strange, I am honestly not sure what this is supposed to accomplish.
I don't think l > slen will ever be true since mbrtowc will only inspect the next slen bytes.
I've removed the casts.

[markj]

lib/libc/locale/collate.c

776

For consistency with other return statements.

811

Shouldn't this return dlen?

821

850

The same odd casting appears here.

865

Is IGNORE_EQUIVAL_CLASS == -1? It's not mentioned in the comment above.

972

[bnovkov]

Address @markj 's comments and fix a couple of style issues.

lib/libc/locale/collate.c

811

Hm, yeah, this doesn't seem to be the correct behavior if the chain is larger than dlen.
I've changed it to return dlen

865

It has a value of 1 and is defined at the top of collate.h.

[markj]

lib/libc/locale/collate.c

988