Re: int21 handling

[Alexandre Julliard]

Eric Pouech  writes:
> I'd suggest changing the oem_files_api to a krnl_file_api_codepage (or
> something like this) which would be set to either CP_ANSI or CP_OEM
> and use this variable in most of file related A functions (instead of
> our current CP_ANSI). Then we would set OEM APIs on int21 entry (and
> reset to the old mode on exit). I assume we hold a lock while
> executing the int21 API, so preventing some multi-threading issues on
> this.

I don't think we want to switch the global flag on every int21 call,
that's fairly ugly IMO. If int21 needs OEM then it should use OEM
explicitly, which is pretty much what it's doing already AFAICS. I see
no reason to change that.