Stephen Leake <[EMAIL PROTECTED]> writes:

> Thus it would seem reasonable that _all_ dispatching functions that
> are interactive should be in dvc-back-end-wrappers.

Why?

A function which is defined with `define-dvc-unified-command' doesn't
need that, since by definition, there is already a <back-end>-whatever
command for each back-end.

`dvc-back-end-wrappers' is only needed when an interactive function is
defined with (defun dvc-<command> ...), with dvc-<command> actually
doing some dispatching later on.

I've put the ones I found in the current definition, can you point out
the missing ones you're thinking of?

> This could prove troublesome; dvc-back-end-wrappers has the argument
> list as well as the name, so any change to the actual function's
> argument list must also be made in dvc-back-end-wrappers. It would be
> nice to be able to check or enforce that automatically, but I suspect
> that's not possible.

Never say impossible ;-). But that's at least non-trivial: the issue
is that at the moment, we can't rely on the loading order. The
dvc-back-end-wrappers stuff can be called before or after the rest of
DVC is loaded (before for bzr*.el, since b comes before d, but after
for xgit*.el for example).

> We should at least put a comment at each function definition that
> says "also update dvc-back-end-wrappers".

Good idea.

-- 
Matthieu

_______________________________________________
Dvc-dev mailing list
[email protected]
https://mail.gna.org/listinfo/dvc-dev

Reply via email to