This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: your recent regclass.c:globalize_reg change
- To: Joern Rennecke <amylaar at cygnus dot co dot uk>
- Subject: Re: your recent regclass.c:globalize_reg change
- From: Jeffrey A Law <law at cygnus dot com>
- Date: Sun, 28 Nov 1999 06:05:29 -0700
- cc: aph at cygnus dot co dot uk (Andrew Haley), gcc-patches at gcc dot gnu dot org
- Reply-To: law at cygnus dot com
In message <199911252308.XAA20975@phal.cygnus.co.uk>you write:
> When you moved the error (it is indeed an error, not a warning as you
> wrote in the ChangeLog) for global registers being defined after too
> late from make_decl_rtl to globalize_reg, you removed the check
> that allows fixed registers to be declared as register variables at
> any time. This breaks newlib/libc/sys/sh/syscalls.c:
>
> ...
> int
> _close (int file)
> {
> return __trap34 (SYS_close, file, 0, 0);
> }
>
>
> register char *stack_ptr asm ("r15");
>
> caddr_t
> _sbrk (int incr)
> ...
>
> Declaring a fixed register as a variable can be done at any time, so the
> check to allow this should be preserved.
>
> Mon Nov 22 23:53:50 1999 J"orn Rennecke <amylaar@cygnus.co.uk>
>
> * regclass.c (globalize_reg): Re-instate test that allows
> fixed registers to be declared as a variable even after functions
> are defined.
Thanks. Please install the testcase into the gcc.dg regression tests
and please install the fix into the mainline sources.
Thanks for also fixing the spelling typos.
jeff