Re: [PATCH] Add Rust support to source highlighting

[Tom Tromey <tom at tromey dot com>]

>>>>> "Pedro" == Pedro Alves <palves@redhat.com> writes:

>> It seems like a bit of a pain to fix at top-level.  These flags are put
>> into the default LDFLAGS for stage1, or when not bootstrapping.

Pedro> Can you expand on why is it a pain?  I was imagining that the top-level
Pedro> script would take in consideration whether a gcc/ subdir exists, in
Pedro> addition to checking some --{enable,disable}-static-runtime or some such,
Pedro> where it adds the flags to LDFLAGS.

The main problem is that the flags are passed down from the top-level
Makefile, so it would need extra top-level Makefile.* hacking.

Pedro> I assume it is put in LDFLAGS for the whole tree in order to
Pedro> use -static-libcc consistently for both gcc and the libraries it
Pedro> depends on (like libiberty).  (It'd be interesting to find the
Pedro> rationale in the original mailing list post/patch that added it to
Pedro> be sure.)

Pedro> With what you're suggesting it sounds like we'd build libiberty/, libbfd/,
Pedro> etc. with -static-libgcc and gdb/ without?  That sounds like something
Pedro> we shouldn't be doing either.

Are those even useful for libiberty or bfd?  I thought those only
affected the link.

Or do people build a shared libiberty and/or bfd?  That seems bad.

I tend to think -static-* is not ever correct for gdb, or at least is
incompatible with Source Highlight.

Tom