Re: CVS commit: src/sys/arch/mips

[Michael <macallan%netbsd.org@localhost>]

Hello,

On Tue, 13 May 2014 08:01:41 +1000
matthew green <mrg%eterna.com.au@localhost> wrote:

> > > >   gcc 4.5, gmp patched as
> > > > posted earlier.
> > > 
> > > Is this ready to commit?
> > 
> > Good question, I'd like to hear from one of the Matts first. All the
> > patch does is to disable using assembly routines when building n32
> > binaries ( as in, mips64* but not _LP64 ). Seems to work, at least on
> > sgimips, needs more torturing on Loongson.
> 
> there are a couple of minor problems from my POV:
> 
> - they're mostly generated files you've changed (at least,
>   the Makefile.inc's..)

I'm thoroughly unfamiliar with gmp, what should I change instead?

> - the comments in the code seems to indicate that what we
>   have right now *should* work, as it was designed to work
>   with irix n32.  mac's experience makes me question this.

That's what matt@ seems to think as well. I wonder if he ever tried to
build anything natively on his mips64 hardware.
Also, isn't this supposed to work on Linux/MIPS? Last time I checked
they were using n32 as default ABI. The real problem is probably
somewhere else and I'm just working around it.

> i think they could be commited right now, but please add
> an entry to doc/HACKS.

Will do, thanks.
The compiler definitely works with this on sgimips ( built a couple
packages like perl, gmake, tcsh with working results, perl's make test
is still running and seems to pass pretty much everything ), I didn't
get around to do more torturing on Loongson yet.
Symptoms were the same though, alignment faults from gmp's assembly
routines doing 64bit ops on 32bit data when using -On with n>0.

have fun
Michael