08:17 <fiddlerwoaroof> luis: I'm still not sure what I should do about this https://github.com/slime/slime/pull/487#discussion_r249446910

08:18 <fiddlerwoaroof> Was you objection using the internal function of the swank-repl package?

08:20 <fiddlerwoaroof> I guess I could copy this lines in, but that isn't very elegant either

08:20 <fiddlerwoaroof> https://github.com/fiddlerwoaroof/slime/blob/output-streams/contrib/swank-repl.lisp#L127-L137

08:21 <fiddlerwoaroof> Maybe I should copy them here: https://github.com/fiddlerwoaroof/slime/blob/output-streams/swank/gray.lisp#L208

08:28 <fiddlerwoaroof> I've also been thinking it would be nice to add a couple features as a follow-up, once I get the minimal version in:

08:30 <fiddlerwoaroof> 1. moving the marker with file-position, I think this would be somewhat useful, especially once we make these bi-directional

08:31 <fiddlerwoaroof> 2. setting the buffer mode: I write a lot of code that generates org-mode files/etc. and it'd be nice to somehow tell emacs which mode is relevant to the file... although, I guess this might be a security issue?

08:31 <fiddlerwoaroof> 3. some kind of dynamic font-lock support: i.e. picking colors, at least, but potentially also picking fonts and font-sizes. I guess we'd need to detect if emacs is running in GUI mode for this?

08:34 <fiddlerwoaroof> ... I'd be a bit inclined to implement the primitives necessary to have a CLIM-style output stream

08:35 <fiddlerwoaroof> anyways, I'm off to bed

23:05 <luis> fiddlerwoaroof: I think swank-buffer-streams shouldn't depend on swank-repl at all and neither should slime-buffer-streams depend on slime-repl on the Emacs side. I think you should write specific buffer management code for the new contrib. If a refactoring opportunity shows up, then we can add some shared code to the main bits of slime/swank.

23:06 <fiddlerwoaroof> I think the function I'm using is really too generic for swank-repl

23:06 <fiddlerwoaroof> So, I'd propose moving it next to swank's implementation of gray streams

23:07 <fiddlerwoaroof> That's the only dependency

23:07 <fiddlerwoaroof> Ah, I guess there's also the hash-table

23:11 <luis> fiddlerwoaroof: for this feature, I think we should use a line-buffered stream. I didn't mention that before because it's probably not needed for the first minimal version, but it might affect how much is actually shared with swank-repl's stream code.

23:12 <fiddlerwoaroof> I've been thinking about that too, as well as emacs performance issues when a single line gets too long

23:12 <luis> fiddlerwoaroof: but, I agree, it seems likely that this stream code belongs in a more generic place

23:12 <fiddlerwoaroof> It would probably be good to do something if a line crosses a certain width

23:13 <luis> fiddlerwoaroof: right, Emacs panics a little a bit with long lines. I heard they're working on improving that.

23:14 <luis> fiddlerwoaroof: presumably the repl would have similar issues, so maybe we can just ignore that issue. :-)

23:14 <luis> 🤞

23:17 <fiddlerwoaroof> I think we could also factor this code out: https://github.com/fiddlerwoaroof/slime/blob/output-streams/contrib/slime-repl.el#L296-L325

23:17 <fiddlerwoaroof> slime-output-target-marker probably shouldn't be factored out completely: rather the repl-specific parts could stay here nad the t case moved into a more generic function

23:23 <luis> Agreed.

23:24 <luis> If you're slow enough to implement this, I might have the swank event dispatching simplification ready in time! ;-)

23:27 <fiddlerwoaroof> do external functions have to be declared through definterface?

23:27 <fiddlerwoaroof> (in the swank package)