[m-rev.] for post-commit review: use actual types in groundness tests

[Zoltan Somogyi]

On 2024-07-03 11:25 +10:00 AEST, "Julien Fischer" <jfischer at opturion.com> wrote:
> On Wed, 3 Jul 2024 at 02:22, Zoltan Somogyi <zoltan.somogyi at runbox.com> wrote:
>>
>>
>> On 2024-07-02 22:22 +10:00 AEST, "Zoltan Somogyi" <zoltan.somogyi at runbox.com> wrote:
>> >
>> > On 2024-07-02 22:20 +10:00 AEST, "Julien Fischer" <jfischer at opturion.com> wrote:
>> >> I haven't had a chance to review this yet, but either this change or
>> >> the following one appear
>> >> to have broken compilation of the library in the
>> >> asm_fast.gc.profdeep.tr.stseg grade in this
>> >> evening's rotd build.
>> >
>> > I will have a look.
>>
>> This diff fixes the bug.
> 
> Thanks for that.

You are welcome. I did add the bug before fixing it :-(

>> My bootcheck in that grade now works
>> with only one failed test case, trailing/test_trail_reset. Since I don't usually
>> bootcheck in trailing grade, I don't know whether that is normal.
> 
> Failures in that directory in non-profdeep trailing grades are not normal.
> With rotd-2024-07-01, test_trail_reset seem appear to be invoking the function
> trail entries more times than in the non-profdeep cases; if that's what you are
> observing then it's nothing caused by this change.

The diff was this:

--- test_trail_reset.exp        2023-09-03 15:09:17.454911968 +1000
+++ test_trail_reset.out        2024-07-03 01:39:32.463771416 +1000
@@ -1,3 +1,7 @@
+calling function trail entry with 10: commit
+calling function trail entry with 11: commit
+calling function trail entry with 11: commit
+calling function trail entry with 10: commit
 calling function trail entry with 11: commit
 calling function trail entry with 10: commit
 Call to test/1 succeeded

Note that deep profiling does add the Mercury equivalents of
choice points in some cases to get control when a predicate
is about to fail. If trailing does not know about these, and I don't think
it does, then that may explain some extra trail activity. I don't know
whether it explains exactly *this* trail activity.

Zoltan.