[PATCH 4/4] Revert "Arm: correct macro use in gas testsuite"

[Fangrui Song]

On Fri, Aug 16, 2024 at 5:35 AM H.J. Lu <hjl.tools@gmail.com> wrote:
>
> On Fri, Aug 16, 2024 at 5:31 AM Jan Beulich <jbeulich@suse.com> wrote:
> >
> > On 16.08.2024 13:33, H.J. Lu wrote:
> > > On Fri, Aug 16, 2024 at 4:23 AM Jan Beulich <jbeulich@suse.com> wrote:
> > >>
> > >> On 14.08.2024 18:54, H.J. Lu wrote:
> > >>> This reverts commit cfa18744d435b55bbbbc5ef1ae1df67e84aa1777.
> > >>>
> > >>> commit 6ae8a30d44f016cafb46a75843b5109316eb1996
> > >>> Author: Jan Beulich <jbeulich@suse.com>
> > >>> Date:   Fri Aug 9 11:59:31 2024 +0200
> > >>>
> > >>>     gas: have scrubber retain more whitespace
> > >>>
> > >>> has been reverted to fix PR gas/32073.
> > >>
> > >> This is the one I said should rather stay in any event. Even for the
> > >> other three (plus a fourth one that you didn't revert) my earlier
> > >> indication was that they _may_ want reverting, not that they need to
> > >> be reverted right away. I think we first want to have a clear picture
> > >> on what the new behavior is going to be. After all it's quite
> > >> possible that for ambiguous cases we will warn people (e.g. telling
> > >> to better switch to comma-separated arguments).
> > >>
> > >> Plus, if I may ask: Who did approve these going in? I think reverts
> > >> can hardly ever count as "obvious".
> > >
> > > Many assembler testcases came from real usages.
> >
> > Especially on Arm (32-bit and 64-bit) afaict many macro uses are to reduce
> > the volume of the testcases, not because respective code would be found in
> > "real" use cases.
> >
> > >  Make working
> > > testcases as error will potentially break existing applications.   These
> > > tests shouldn't be changed to begin with.
> >
> > This Arm one wasn't working right, as explained in the description of that
> > patch. I don't think we should deliberately test broken uses of macros.
> >
> > Jan
>
> That is your opinion.  I don't think we should change working tests to error
> since we don't know if they came from real applications.
>
>
> --
> H.J.

I have written down my thoughts on
https://sourceware.org/pipermail/binutils/2024-August/136322.html

This particular "ldrtest2 \load \sym \offset" usage is definitely weird.

   ldrtest ldrd strd f "+ 256" => ldrtest ldrd f + 256   // treat f +
256 as one single argument

I don't think changing the usage will cause any disruption...

I cannot find space-separated macro calls outside the binutils testsuite.
Perhaps we can deprecate space-separated arguments as deprecated and
report warnings.