Re: [AC] help synced with FSF and removale of button-2
"Stephen J. Turnbull" <[email protected]>
| Newsgroups | gmane.emacs.xemacs.patches |
|---|---|
| Message-ID | <[email protected]> |
Mats Lidell writes: > >>>>> Stephen J Turnbull <[email protected]> writes: > > Thanks for the feedback! > > > You've also deleted some functions without saying so, and without > > saying why. That's a no-no, somebody might have code in init.el that > > uses them. > > >From the top of my head I think it was a rename due to the completely > new way to handle stuff. Right. This is a complete API break, it needs to be described that way. > Telling why I changed it will not help people with it in init.el. It won't fix the breakage you cause, but at least they'll understand what happened (or feel sheepish when you tell them it's in the changelog). Also you can make it clear there's a new function that does pretty much the same thing. This is especially important when the change isn't one-to-one (I think you remove two functions and replace them with one general one?) A more complex strategy is to keep the old code and make sure nothing in core or packages uses it. Then use `make-obsolete' to inform users (although they may not compile their inits, in which case they won't see a warning from the compiler). > What is your suggestion? Keep the old name although everything is > different is of course an easy way out. No, you're doing the right thing. Either you keep the old code or reimplement it with pretty close to the same semantics using the new code, or you just break things, and hope there's not much of that. In this case, there probably is not -- I would say just do the sync, document it, and hope. > >> --- a/tests/automated/keymap-tests.el Mon Jan 27 17:52:33 2014 +0100 > >> +++ b/tests/automated/keymap-tests.el Sun Mar 09 00:36:29 2014 +0100 > >> @@ -36,7 +36,7 @@ > >> find-function-at-point Q help-mode-bury button2 > >> help-mouse-find-source-or-track p > > A > > Have you run the tests? Doesn't look like it --+ > > I did in october and that is why keymap-test.el was updated since they > broke then. I will have look again. Yeah, I'm just saying that if the test actually tries to call `help-mouse-find-source-or-track', it should fail loudly.