bug#69602: 29.1; Image :map should adjust with :scale and :rotation

[Eli Zaretskii]

> From: Joseph Turner <joseph@breatheoutbreathe.in>
> Cc: 69602@debbugs.gnu.org, stephen.berman@gmx.net, juri@linkov.net
> Date: Fri, 22 Mar 2024 17:11:17 -0700
> 
> > Thanks.  The tests you added have some problems:
> >
> >  . you use thread-first, but don't require subr-x when compiling
> 
> I removed thread-first.
> 
> >  . the tests fail when run in batch mode
> 
> I added (skip-unless (display-images-p)) to the two problematic tests,
> and it solved the issue on my machine.

Solves it here as well.

> >  . when invoked interactively in a GUI session, one test fails:
> >
> >     F image--compute-map-and-original-map
> >     Test ‘image--compute-map’ and ‘image--compute-original-map’.
> >     (ert-test-failed
> >      ((should (equal (image--compute-map image) flipped-map)) :form
> >       (equal
> >        (((circle ... . 24) "a" (help-echo "A"))
> >         ((rect ... 162 . 149) "b" (help-echo "B"))
> >         ((poly . [6 29 7 22 13 15 21 10 31 7 ...]) "c" (help-echo "C")))
> >        (((circle ... . 24) "a" (help-echo "A"))
> >         ((rect ... 161 . 149) "b" (help-echo "B"))
> >         ((poly . [5 29 6 22 12 15 20 10 30 7 ...]) "c" (help-echo "C"))))
> >       :value nil :explanation (list-elt 0 (list-elt 0 (cdr (car ...))))))
> >
> > It looks like some pixels do not match exactly?  Perhaps some
> > tolerances need to be allowed?
> 
> Interesting - does the result of `image-size` vary per machine?

I guess so.  The transformations are AFAIK done in floating-point
arithmetics, so some minor inaccuracies are possible.

> In any case, I added `image-tests--map-equal' to compare image maps with
> some tolerance. Do the tests pass on your machine now?

Yes, they do now, thanks.  However, there's a warning when compiling
the tests:

  In image-tests--map-equal:
  lisp/image-tests.el:192:17: Warning: Unused lexical variable `i'

Can you fix this, please?