Re: [PATCH] Reallow unrestricted use od .set mipsX

[Thiemo Seufer <ica2_ts at csv dot ica dot uni-stuttgart dot de>]

Richard Sandiford wrote:
> Thiemo Seufer <ica2_ts@csv.ica.uni-stuttgart.de> writes:
> > this patch re-allows to use .set mipsX pseudo-op without getting
> > hindered by an -mgp32/-mfp32/-mabi=32 option, which currently
> > restrict register size and usable insn's over the whole file.
> > 
> > This restores the old behaviour, which is also the way
> > traditional mips assemblers work. Currently e.g. a
> > 
> > 	.set mips3
> > 
> > in a -mabi=32 file is rendered useless.
> 
> Sorry to be complaining again!
> 
> I've no objections to the -mabi=32 part, especially if that's what
> traditional MIPS assemblers do.  But -mfp32 and -mgp32 are GAS's
> options, aren't they?  They never used to be affected by .set mipsX
> options,

"Never" is a strong word for something which was implemented a few
weeks ago. :-)

>and I'm not sure it makes sense for them to be.

.set mipsX is normally used to loosen a restriction, otherwise
there's no point in having it. It's not intended for usual
programming but for stuff like exception handlers, and
traditionally there are no guarantees made about it to the
average programmer execpt of being blamed when it doesn't work.


Thiemo