Re: recent emacs font problems?

[Richard Stallman <rms@gnu.org>]

    Reading through the case and the comments, and comparing this to the
    code, I'm uncertain whether the intention of the fix is to _only_
    avoid overriding the :family property (that's not what the fix does,
    but what the comment below says), or whether the intention really is
    to avoid overriding any of the properties (as the fix does).

I think the old bug would have occured for any of the attributes that
are part of the font name.  But I did not test others, only :family.

I would guess that the attributes for which the new bug occurs are the
same ones for which the old bug, that was fixed by my change, used to
occur.  I suspect that the correct treatment is the same for all these
attributes.  But I don't know what is correct.  Someone needs to
figure out what happens in the case that fails now, why the change
caused this new bug, and how to distinguish the case in which we want
to forcibly set all the attributes there from the case in which we
don't.