This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [tree-ssa] Merge status



On Mar 14, 2004, at 10:04, Joseph S. Myers wrote:
If SPEC uses extended lvalues, how did it ever work with non-GCC
compilers?

Actually SPEC does not use extended lvalue for compilers other than GCC because of the
way obstack.h was written. For some reason nobody say that obstack.h was helping out
GCC but hurting other compilers.
In fact Joeseph, you are the one who had obstack.h fixed:
2003-10-22 Joseph S. Myers <jsm@polyomino.org.uk>


* obstack.h: Merge the following change from gnulib:
2003-10-21 Paul Eggert <eggert@twinsun.com>
* obstack.h (obstack_1grow_fast): Properly parenthesize arg.
(obstack_ptr_grow_fast, obstack_int_grow_fast):
Don't use lvalue casts, as GCC plans to remove support for them
in GCC 3.5. Reported by Joseph S. Myers. This bug
was also present in the non-GCC version, indicating that this
code had always been buggy and had never been widely used.
(obstack_1grow, obstack_ptr_grow, obstack_int_grow, obstack_blank):
Use the fast variant of each macro, rather than copying the
definiens of the fast variant; that way, we'll be more likely to
catch future bugs in the fast variants.


In fact there are other uses of extensions in SPEC CPU 2000 in 176.gcc.
Lets hope they can get this right for SPEC CPU 2004.

Thanks,
Andrew Pinski


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]