Stream: t-compiler

Topic: pre-steering meeting 2019.05.03


nikomatsakis (Apr 30 2019 at 17:01, on Zulip):

Hi @T-compiler/meeting! It's been 3 weeks since our last steering meeting, in which we discussed the idea of a design meeting. I recently landed a first draft of a more specific proposal, which you can find on the compiler-team repository.

The purpose of this topic is to discuss what we should do in this week's steering meeting (this Friday). I wanted to suggest that we review the proposed design meeting process:

This may not take the full time. That seems ok, we can always cancel, but people might also have other things they'd like to talk about to fill the time.

Anyway, does this sound good? Other thoughts?

eddyb (Apr 30 2019 at 17:07, on Zulip):

maybe I should write up a proposal for https://github.com/rust-lang/rust/issues/43081#issuecomment-487985457 before getting into all the "def layer" stuff?

eddyb (Apr 30 2019 at 17:08, on Zulip):

it seems more pressing (as anything that happens to slip for months :P)

eddyb (Apr 30 2019 at 17:13, on Zulip):

it will probably involve enough changes to allow macro expansion to become incremental (modulo name resolution, I suppose)

nikomatsakis (Apr 30 2019 at 20:03, on Zulip):

maybe I should write up a proposal for https://github.com/rust-lang/rust/issues/43081#issuecomment-487985457 before getting into all the "def layer" stuff?

@eddyb I was actually going to suggest exactly that. This feels like a topic where it'd be great to spend a bit of time to (a) bring it back in cache and (b) talk about the right sol'n

eddyb (May 01 2019 at 09:18, on Zulip):

I think I know how to tie it into @matklad's ideas for the future of parsing

eddyb (May 01 2019 at 09:19, on Zulip):

by layering the problems and picking out what we can do now to improve the proc macro situation

eddyb (May 01 2019 at 09:28, on Zulip):

@nikomatsakis one thing I'm struggling with is where to put a very-long-term plan

eddyb (May 01 2019 at 09:28, on Zulip):

which we may want to discuss bits of

eddyb (May 01 2019 at 09:29, on Zulip):

I guess I'll use hackmd and then pick parts from it

matklad (May 01 2019 at 10:24, on Zulip):

Thanks for cc eddyb ! This is indeed an interesting problem which has implications for IDEs, so I am especially interested in the long-term vision here. I have ideas about how to reconcile token-tree based AST, which is needed for macros, and CST-based AST, which is arguably needed for IDEs, but I haven't actually gotten to trying them out.

I also don't have a usual class this Friday, so I'll be available for the steering meeting.

Last update: Nov 16 2019 at 01:05UTC