On 9 Feb 2007, at 22:30, Kim F. Storm wrote:
It does not apply to just self-insert-command. What if you also
created slime-kill-line, slime-yank, slime-next-word, etc. None of
these would call self-insert-command, but they would still have to be
marked suitably to interact seamlessly with CUA and d-s-m.
I see.
The problem is that if the writer of slime mode is not aware of the
necessity to interact with CUA or delete-selection-mode, documenting
the interface wouldn't really help (until some user complains of stuff
not working with e.g. CUA)...
To make people aware, there should be a section detailing what it
takes to implement interactive commands that operate on a buffer.
Maybe this already exists somewhere...
To sum up:
- You require that every externally supplied interactive command
needs to be encoded to work with CUA and d-s-m, with a sensible
default set.
- This also means that replacing `self-insert-command' with one's own
function in some keymap and calling self-insert-command from that
function will not yield the same result. This is counter-intuitive.
It doesn't seem like we can come up with a technical solution. I
don't want to keep anyone from working on the release, so I'll shut
up here. I suggest proper documentation of this class of cases though.
D
_______________________________________________
emacs-pretest-bug mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/emacs-pretest-bug
