Helmut Eller e9626484@stud3.tuwien.ac.at writes:
Which versions of the various Lisp to we want to support? The latest released version in June?
That would be nice, but I think we should keep the backwards compatibility that we have if people are need it. We're testing with ACL5.0 so we should stay compatible with that if possible, and I think people are using older LispWorksen out of necessity (please confirm!).
Raymond Toy also suggested we shouldn't be too quick to drop CMU18e support, since 19a will want some field testing before it's considered "the right stuff". But 19a won't be out by end of June anyway, but probably will by August. I would love to delete the 18e-compat code but I think we should wait until (immediately) after 1.0.
For SBCL, OpenMCL, CLISP I don't know of any reason to aim for compatibility with older versions.
We should also try to make sure that we'll be forwards-compatible for a while by minimizing our uses of internal functions and hassling implementors not to delete the ones that we still use too quickly :-)
The SBCL backend definitely needs an audit since those guys are so good about making supported interfaces for us.
Tests for the "essential" features we be nice. It would also be nice if we could write tests for a specific Lisp.
Yes. In addition to the fully automated tests it might be good to write an interactive tester that sets up certain situations (file with compiler notes, sldb, etc) and asks you to check certain things manually. I'm not sure. It was helpful in Distel, but then I didn't have any fully-automated tests there.
Are there any features we can remove? I was going to suggest removing the REPL (because the code is messy and a REPL is not the Emacs way to interact with anything), but I guess people wouldn't like that :-)
Amazingly I find myself using the REPL all the time, even though I never use IELM for Emacs Lisp. I'm not sure why.
Any outstanding keybinding wars? A while ago we talked about grouping documentation commands under C-c C-d; I like that idea.
Agree. We should also move hyperspec-lookup so that `C-c C-h' does the default action of listing all bindings starting with C-c, and move the funky-indentation C-M-q command because it's different to the standard indentation command on that binding (does a whole defun, not sexp at point).
Should we make a (final?) try to simplify the connection handling code in swank.lisp? It is not very readable, but I haven't any good ideas to improve it.
Lots of code-cleanups would be good.
I'm still tempted to delete all of the multiple-connections code. It seems to me like the only thing we use multiple-connections-to-one-lisp for is attaching a new connection to a thread, but that seems multiplexable in principle at least. I'll reread what we said about this last time.
-Luke