bug#41544: 26.3; Possible incorrect results from color-distance

[Mattias Engdegård]

20 aug. 2020 kl. 15.08 skrev Lars Ingebrigtsen <larsi@gnus.org>:

> Mattias?

I don't want to revive this discussion and consider the matter settled, but 
since you asked:

Suppose we add standard-colour-darkness-predicate, say, with color-dark-p being 
the default and (r+b+g)/3<0.6 the 'traditional' option (or the default, if you 
prefer). How would the effects of that choice be explained to the user?

A difference is only visible when:

1. a face sensitive to the background mode is used (many standard faces are), 
and
2. a background colour used that is judged differently by the available 
predicates

Most reasonable backgrounds are either too light or too dark to pass the latter 
criterion, but people's idea of what is reasonable varies. For instance, 
black-on-green text is somewhat readable, but

  emacs -bg green -fg black

will give mostly bad default faces (for instance, the minibuffer prompt is 
almost invisible). With color-dark-p it becomes workable (attached diff).

Whether this is reason enough to add a defcustom is something I won't comment 
on. I doubt anyone will change it, whatever the default would be.

dark.diff
Description: Binary data