This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Patch to move "#define USG" from xm-*.h to config.gcc's xm_defines
- To: "Kaveh R. Ghazi" <ghazi at caip dot rutgers dot edu>
- Subject: Re: Patch to move "#define USG" from xm-*.h to config.gcc's xm_defines
- From: "Zack Weinberg" <zackw at Stanford dot EDU>
- Date: Fri, 9 Mar 2001 01:44:06 -0800
- Cc: gcc-patches at gcc dot gnu dot org
- References: <200103080022.TAA12494@caip.rutgers.edu>
On Wed, Mar 07, 2001 at 07:22:30PM -0500, Kaveh R. Ghazi wrote:
> > From: "Zack Weinberg" <zackw@Stanford.EDU>
> >
> > Are you going to do anything about all the files that just include
> > other files? We want to turn those into direct references from
> > config.gcc.
> > zw
>
> Sure. I actually already took a look but it seemed like it would be
> better to wait until after your config.h patch went in.
>
> I'm going to hit POSIX next.
> After that I'm going to put MAXPATHLEN into system.h.
> Then I'll wait for your patch to go in and do include files.
>
> Besides those and:
> HOST_WORDS_BIG_ENDIAN
very next on my list
> ONLY_INT_FIELDS
> HOST_FLOAT_FORMAT
for autoconf to handle these will require great cleverness; not
impossible, though. ONLY_INT_FIELDS - we have to think of tests for
all the possible ways compilers could mess up enum bitfields.
HOST_FLOAT_FORMAT - I imagine a hack like rth's for
HOST_WORDS_BIG_ENDIAN would work, but we'd have to find floating point
numbers that correspond to recognizable strings that could be grepped
for. That's not easy.
> Deleting USG/POSIX from config.gcc
For this to happen, changes have to be made in the debug output code
as well as the rest of the compiler.
> Bypassing platforms which really need hacks: vms/beos/dos/interix/etc
I'm pretty sure beos doesn't need most of what it's got now ... it can
run configure just fine.
> What else is on the "list"?
Sorting out things that belong in target headers but are in host.
Interix, VMS, DOS, etc. have a bunch of this.
The x-host files. All kinds of random crap hiding in there. Most of
it totally obsolete, I think.
Clean up pathname handling to the point where all we have to do is
define HAVE_DOS_BASED_FILE_SYSTEM to get the DOS path handling, and
similar for VMS. I think everyone else uses the simple Unix filename
scheme.
Weed out dead configurations: vax-vms and A/UX come to mind.
See how much of the t-target files can go away. Two things in
particular: we should be running fixproto everywhere, just like
fixincludes; we should stop providing our own @#$! assert.h except on
platforms which really don't have it.
zw