Re: CUA/delete-selection-mode and remapped keys incompatibility

[David Reitter]

On 9 Feb 2007, at 15:12, Stefan Monnier wrote:

> The implementation I'm imagining would be basically as follows:
> cua-mode (and delete-selection-mode) add a (global) before-change- 
> function
> which outputs this warning.  This would be modular and catch all  
> cases I can
> think of, but maybe it would catch more than we'd want.

Why do you want to bother the end user with such a warning?

As of now, the developer of a package such as SLIME would have a look  
at what a key is bound to. They find `self-insert-command' and then  
construct a function that calls this function with the same argument.  
If they are clever, they might even use "call-interactively".

At that point, they believe that self-insert-command when called from  
their function does exactly the same thing as when called (really)  
interactively. This is a reasonable assumption in a functional  
environment. And in fact, people write such code, as demonstrated. (I  
believe the real issue has to do with `last-command'...)

The immediate solution to this is to document the necessary  
annotation of functions in the right place. Where? Maybe along with  
`self-insert-command' - I don't really know.