fix weakref tests for alpha

classic Classic list List threaded Threaded
4 messages Options
Reply | Threaded
Open this post in threaded view
|

fix weakref tests for alpha

Richard Henderson-2
Alpha assembly obsconds with .set for other purposes.
As far as I know, = means the same thing everywhere.

Tested on alpha and i686.


r~


        * gas/all/weakref1.s: Use "=" instead of ".set" for equivalence.

Index: gas/all/weakref1.s
===================================================================
RCS file: /cvs/src/src/gas/testsuite/gas/all/weakref1.s,v
retrieving revision 1.2
diff -u -p -d -r1.2 weakref1.s
--- gas/all/weakref1.s 4 Nov 2005 19:45:25 -0000 1.2
+++ gas/all/weakref1.s 16 Nov 2005 21:59:38 -0000
@@ -189,7 +189,7 @@ l:
 /* d# target symbol definitions */
  .weakref Wld1, ld1
  .long Wld1
- .set ld1, l
+ld1 = l
 
  .weakref Wld2, ld2
  .long Wld2
Reply | Threaded
Open this post in threaded view
|

Re: fix weakref tests for alpha

Alexandre Oliva-2
On Nov 16, 2005, Richard Henderson <[hidden email]> wrote:

> Alpha assembly obsconds with .set for other purposes.
> As far as I know, = means the same thing everywhere.

'fraid not, I'd used it initially, and was asked to change it because
some ports didn't support it.  Maybe .equ or .equiv would give us more
portability?

--
Alexandre Oliva         http://www.lsd.ic.unicamp.br/~oliva/
Red Hat Compiler Engineer   aoliva@{redhat.com, gcc.gnu.org}
Free Software Evangelist  oliva@{lsd.ic.unicamp.br, gnu.org}
Reply | Threaded
Open this post in threaded view
|

Re: fix weakref tests for alpha

Hans-Peter Nilsson-2
On Thu, 17 Nov 2005, Alexandre Oliva wrote:
> On Nov 16, 2005, Richard Henderson <[hidden email]> wrote:
>
> > Alpha assembly obsconds with .set for other purposes.
> > As far as I know, = means the same thing everywhere.
>
> 'fraid not, I'd used it initially, and was asked to change it because
> some ports didn't support it.

(raises hand)
Um, MMIX, where these tests used to PASS...

>  Maybe .equ or .equiv would give us more
> portability?

Looking at read.c, (only) the generic .equ is a synonym for the
generic .set, but tc-hppa.c and tc-tic4x.c redefines them
incompatibly.  Maybe a new pseudo?

If you must, please skip these tests for MMIX.

brgds, H-P
Reply | Threaded
Open this post in threaded view
|

Re: fix weakref tests for alpha

Jan Beulich
I was actually thinking of a different approach to address the .equ/.set
issue in gas/testsuite/gas/all generally: Since all targets, as far as I
was able to figure out, support at least one of generic .set or generic
.equ, why not abstract this out and create an include file which defines
a macro, expanding to either .set or .equ depending on the target. Short
of a gas option to force inclusion of a file from the command line, a
command line option indicating the desired selection would then need to
be passed (perhaps globally, so that other non-target-specific tests
could also benefit).

Also, I got just now aware of Richards change to the test, my checkin
earlier today (but after Richard's) actually disabled the test for all
the targets that are known to implement .set differently (the same
exclusion set as on other tests using .set), so potentially reverting
Richard's change would still leave the alpha (and other affected)
testsuite(s) not failing on these tests anymore.

Jan

>>> Hans-Peter Nilsson <[hidden email]> 17.11.05 10:49:37 >>>
On Thu, 17 Nov 2005, Alexandre Oliva wrote:
> On Nov 16, 2005, Richard Henderson <[hidden email]> wrote:
>
> > Alpha assembly obsconds with .set for other purposes.
> > As far as I know, = means the same thing everywhere.
>
> 'fraid not, I'd used it initially, and was asked to change it
because
> some ports didn't support it.

(raises hand)
Um, MMIX, where these tests used to PASS...

>  Maybe .equ or .equiv would give us more
> portability?

Looking at read.c, (only) the generic .equ is a synonym for the
generic .set, but tc-hppa.c and tc-tic4x.c redefines them
incompatibly.  Maybe a new pseudo?

If you must, please skip these tests for MMIX.

brgds, H-P