Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> The attached patch should fix it.
>
> The patch looks OK tho I can't see why you add `intervals_equal_1` to
> intervals.h (instead of marking it static).

Me neither, fixed in the attached.  (It was an artifact of some earlier
experimentation that shouldn't have survived.)

>> (compare_string_intervals): Call intervals_equal1 with third
>> argument as true.
>
> Rather than paraphrase the code change this could say what's the
> intended effect of passing that arg.

I'm not a fan of the GNU ChangeLog format, as it seems to encourage bad
habits like not useful descriptions and massive patches (as opposed to
patch sets).  In any case, I've made an attempt to force a useful
description into this clunky format in the attached.

I also see that ert had an inefficient re-implementation of
'equal-including-properties' lying around as a workaround for this bug,
so I've fixed that up in a second patch.

I hope that Eli or Lars could say if they prefer the first patch (with
the bug fix) on emacs-28 or master.

The second patch is not urgent and should definitely go to master.