@oli @RalfJ @centril @Christian Poveda @eddyb (sorry)
There's been some developments in const-eval recently, and an official working group would help to communicate these developments to the compiler team as well as the broader community. I've been working on static validation of const bodies to enable #49146, #52000 and (in the distant future) #57349, and others have been working on the MIRI side, including adding checks to ensure that the static validation is sound.
I think this hasn't been done in the past because migrating/creating a repo, scheduling meetings and writing progress summaries is not very fun, and most of the stakeholders have a lot of other responsibilities on the rust project and elsewhere. I'm relatively new here, I'm not a compiler team member and I don't have a good understanding of the inner workings of MIRI, but I would be willing to take on some of these organizational tasks, including generating the weekly report to the compiler team. However, I think @oli and @RalfJ should be the co-leads for this WG.
We should also try to have a weekly scheduled meeting.
I think we will want this to be a joint t-lang & t-compiler effort
@oli should be the t-compiler liason and I can be that from t-lang
but I'll probably not do any leading wrt. dataflow because I need to read up on that :D
don't think we have to rename the zulip stream tho or such; we can just use this stream as the wg's stream
Well I'll be happy to help in whatever I can
What do you think about migrating
rust-lang/const-eval? I often want to ping a github org instead of individual people (since I inevitably randomly leave someone out), but I think the org doesn't matter for this?
sounds like a good idea; cc @simulacrum
I did not know anything about an rfcs org prior to today
If you transfer the repo to me I can transfer it to rust-lang.
(please also ping me here)
@T-compiler/WG-meta If the other members of
const-eval agree to this, should I open a PR to add
wg-const-eval under rust-lang/compiler-team, rust-lang/lang-team or both? IMO it's closer to the existing
t-compiler WGs than the
t-lang ones, but fits under either umbrella.
let's go with t-compiler but just note that it is a joint effort (in the readme)
See rust-lang/compiler-team/#201. I decided to use "Compile-time Function Evaluation" as the long name for the WG, but this may not be everyone's first choice.
I'd be happy to be involved but don't have time to take a lead role, I am afraid
Actually the way things look right now, UCG+Miri together already take up all the time I have and I probably should not be involved in the const-eval thing :(
I can't even give you an ETA for when I will be able to look at @ecstatic-morse's PR in the const-eval repo; if UCG+Miri activity stays like it is that could take weeks :(
I don't have a strong opinion on where the const eval "rfc" repo lives
making this an official working group is a good idea
I don't know anything about the administrative side of github. I thought having it in
rust-lang would make it a bit more visible.
Also, if we can agree on a short name (
wg-ctfe) I'd like to add it as a team to the rust org.
maybe just wg-const (since const fn is not necessarily just about const eval)
I seem to remember there being a weirdly heated thread about the acronym CTFE in the distant past
(but bikeshed matters not)
Add yourself if you wanna get pinged for big PRs/new issues I guess