Re: [PATCH] Fix BUILTINS_LOCATION with -fpreprocessed (PR preprocessor/41543)

[Tom Tromey <tromey at redhat dot com>]

>>>>> "Jakub" == Jakub Jelinek <jakub@redhat.com> writes:

Jakub> I've kept this so that -E keeps printing:
Jakub> # 1 "test.c"
Jakub> # 1 "<built-in>"
Jakub> # 1 "<command-line>"
Jakub> # 1 "test.c"
Jakub> as it always used to.  I guess at least some programs parsing -E output
Jakub> might be confused if that <built-in> line goes away.

I see, thanks.

Do we know for sure that any code assigning a location to a builtin uses
BUILTINS_LOCATION and not input_location?

Jakub> Well, libcpp side doesn't do anything with UNKNOWN_LOCATION or
Jakub> BUILTINS_LOCATION, all it does is reserve some number of lowest
Jakub> numbered source_location values for caller's special uses.  So,
Jakub> rather than putting those into libcpp libcpp perhaps should
Jakub> define RESERVED_LOCATION_COUNT or something similar, and gcc
Jakub> should assert that BUILTINS_LOCATION is smaller than that.

This sounds good to me.

Tom