That being said, it seems to me that this slowdown is avoidable. Lars
considers that the minor modification he made should not have changed
anything.
Would it be possible for you to run the benchmarks on the current
trunk, with and without that "!"? Because I just don't understand why
disabling the image cache would make Emacs slower.
Sure, I'll do that.
Could you also try with this patch:
@@ -2351,7 +2356,7 @@ lookup_image (struct frame *f, Lisp_Object spec, int
face_id)
/* Look up SPEC in the hash table of the image cache. */
hash = sxhash (spec);
- img = search_image_cache (f, spec, hash, foreground, background, true);
+ img = search_image_cache (f, spec, hash, foreground, background, false);
if (img && img->load_failed_p)
{
free_image (f, img);
It's a definite mistake on my part and may affect the caching of images,
although I can't see any way that it would slow anything down.