This is the mail archive of the mailing list for the binutils 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: PATCH: Fix mixing MIPS object files.

On Tue, Dec 03, 2002 at 04:33:37PM +0000, Richard Sandiford wrote:
> <hangs head>  Sorry for dropping the ball...
> "H. J. Lu" <> writes:
> > Now the question is
> > 
> > # mipsel-linux-as -mips[3|4|5|64]
> > # mipsel-linux-as -march=xxx
> > # mipsel-linux-gcc -mips[3|4|5|64] 
> > # mipsel-linux-gcc -march=xxx
> > 
> > will ever use 64bit registers. I hope they won't. If it is the case,
> > I have no problems to ignore EF_MIPS_ARCH when EF_MIPS_MACH is not
> > zero.
> Here's the sort of patch I had in mind.  Backtracking
> from what I said before, it seems like it's easy to keep
> mips2/sb1-style combinations after all.
> The patch shows no regressions on mips64-elf and makes it
> possible to link 32-bit MIPS IV code with MIPS II code.
> It also behaves the way I expected when running some tests
> by hand.

Have you tried mips-elf?

> It's just a request for comments at this stage.   I'll whip
> up something for the testsuite before seeking approval.

Last time when I tried, I got

# /export/tools-3.2-redhat-8/bin/mipsel-linux-gcc -march=mips64 f.c  -S  
cc1: bad value (mips64) for -march= switch


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