bug#74725: 31.0.50; image-scaling-factor is ignored by create-image

[Eli Zaretskii <eliz@gnu.org>]

> From: Po Lu <luangruo@yahoo.com>
> Cc: da_vid@orange.fr,  alan@idiocy.org,  74725@debbugs.gnu.org
> Date: Sun, 08 Dec 2024 16:03:28 +0800
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> >> > Po Lu, what were the reasons for that particular part of the commit?
> >> 
> >> The scale applied by image-scaling-factor is liable to differ by
> >> display
> >
> > How so?  Please elaborate.
> 
> When it is set to `auto' (the default value), the scaling factor to be
> applied is decided by the configuration of a frame, namely, its
> FRAME_COLUMN_WIDTH.

So when the default font changes, all the images are supposed to be
resized?  Does that really happen, and if so, is that a good idea in
all cases?

> >> and computing the default scale in Lisp would result in images
> >> being displayed with an incorrect scale in the presence of multiple
> >> displays.
> >
> > How does the above changeset solve this problem, then?
> 
> By moving its application to image.c, which knows where an image is
> being displayed and can apply specific scales for each frame.

But, as this bug seems to indicate, that solution doesn't always work?

> >> Image caches must be flushed when image-scaling-factor is modified,
> >> unless it is set to `auto' and a display's scale changes, because
> >> image.c has no means of detecting variable modifications and so only the
> >> latter event can be automatically detected.
> >
> > Please describe the issue in more detail, as I don't think I follow
> > what you are saying here.  If we need to detect changes in variables,
> > we can use the add-variable-watcher technique, similar to what we do
> > in frame.el with variables that need to force redisplay (but maybe I
> > don't understand the problem you are describing).
> >
> > In any case, I don't think changes in image-scaling-factor are
> > supposed to be immediately reflected on display, if that's what you
> > have in mind.  This is not the documented effect of this variable.
> 
> What I am trying to communicate is that changes to
> `image-scaling-factor' must be accompanied by flushing the image cache
> if it is to take effect on all previously displayed images.  This isn't
> a problem, and the OP should simply flush the image cache after
> modifying image-scaling-factor, rather than rely on the erroneous
> behavior of find-image which was removed.

OK, so do you consider the solution of recording the scale factor in
the cache a reasonable one?