[m-rev.] for post-commit review: fix recent bootcheck breakage

[Peter Wang]

On Fri, 24 Jul 2026 02:06:31 +0200 "Zoltan Somogyi" <zoltan.somogyi at runbox.com> wrote:
> For review by anyone.
> 
> The INLINE1 and INLINE files show the output of the new diagnostics
> in inlining.m for the affected predicate in make_hlds_passes.m
> before and after the fix. INLINE1 shows that the original, unsorted lists
> of bound_functors on the coerce goal leads recompute_instmap_delta.m
> to screw up (not surprising, since it is given invalid input), while INLINE
> shows that with the sorted list, this does not happen.
> 
> Zoltan.

> Sort bound functors in the modes of coerce goals.
> 
> This fixes the cause of the breakage of -O5 --intermod-opt bootchecks
> in the asm_fast.gc grade.
> 
> The fix is just the change to modeecheck_coerce.m. Most of the rest
> of the diff contains the changes needed for the detective work
> that led to the fix.

That looks good.

Peter