This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: MIPS port and fixincludes
- To: bkorb at sco dot COM
- Subject: Re: MIPS port and fixincludes
- From: Mark Mitchell <mark at codesourcery dot com>
- Date: Tue, 18 Jan 2000 10:23:07 -0800
- Cc: amylaar at cygnus dot co dot uk, gavin at cygnus dot com, zack at wolery dot cumb dot org, gcc at gcc dot gnu dot org
- Organization: CodeSourcery, LLC
- References: <200001181634.QAA15775@phal.cygnus.co.uk><38849E0B.A955E56F@sco.com>
>>>>> "Bruce" == Bruce Korb <bkorb@sco.COM> writes:
Bruce> Joern Rennecke wrote:
>> > [[...]] For example, it is > pointless to change "sparc" to
>> "__sparc__" on an IA86 box.
>>
>> No it isn't. The user is free to define sparc. (S)he
>> shouldn't get sparc-specific code on an IA86 box as a result.
Bruce> Good argument. The only minor issue is consistency.
I'm going to jump in the middle of an argument that I a) haven't been
following and b) don't know a whole lot about. But, here goes.
I think `fixincludes' should do just the minimum required to get GCC
to be able to *process* the system headers. In other words, if the
system `stdio.h' contains a syntax error (from GCC's point of view),
then fix it. If we need to insert varargs magic somewhere, then fix
it. Etc.
Otherwise, leave the headers alone. If they're broken, they're broken
-- but I've been burned more than once by GCC fixing headers that were
broken. (Like if the software being compiled *expected* the
brokenness, or if something about the fix wasn't quite complete
enough.)
I think it's confusing to have "secret" copies of headers; we should
keep that to a minimum.
--
Mark Mitchell mark@codesourcery.com
CodeSourcery, LLC http://www.codesourcery.com