Stream: t-compiler/wg-polonius

Topic: meeting 2019.10.15


Albin Stjerna (Oct 15 2019 at 08:30, on Zulip):

Hi! :wave: I slept perhaps three hours tonight and will probably have gone to bed (or at least, need to wind down) during tonight's (your today's) meeting, so I'll report on what I have been doing now instead. I have worked on initialisation, including cleaning it up and implementing move error reporting in Polonius. I opened a PR (polonius#135), which contains both a list of things left to do and a high-level description of what I did.

Albin Stjerna (Oct 15 2019 at 08:35, on Zulip):

I guess the thing I'd want some feedback on is: a) is this completely out of whack, and b) how should we proceed to implement move error reporting in Polonius?

lqd (Oct 15 2019 at 08:45, on Zulip):

cool !

lqd (Oct 15 2019 at 19:02, on Zulip):

hello :wave: :)

lqd (Oct 15 2019 at 19:04, on Zulip):

(Albin mentioned at the top of the topic they probably wouldn't be here tonight)

lqd (Oct 15 2019 at 19:04, on Zulip):

a lot of things have happened since last week and I fear I don't remember them all :)

lqd (Oct 15 2019 at 19:05, on Zulip):

in bulk: I've opened a couple PRs, one to fix clap, one as a proposal we might have discussed tonight, or later

lqd (Oct 15 2019 at 19:06, on Zulip):

I've "fixed" the equality prototype

lqd (Oct 15 2019 at 19:06, on Zulip):

a couple of weird things remained in the rustc tests tho

lqd (Oct 15 2019 at 19:06, on Zulip):

one of which I could use your opinion on Matthew, but maybe only Niko knows

Matthew Jasper (Oct 15 2019 at 19:07, on Zulip):

Do you have a link?

lqd (Oct 15 2019 at 19:08, on Zulip):

(all this is in the dedicated thread and I was wondering if the infinite loops should compile or not)

lqd (Oct 15 2019 at 19:08, on Zulip):

in the https://github.com/rust-lang/rust/blob/master/src/test/ui/borrowck/mut-borrow-in-loop.rs case

lqd (Oct 15 2019 at 19:10, on Zulip):

I wondered about the FactTypes thing as well, I'm a bit unsure it's worth it in the end

lqd (Oct 15 2019 at 19:10, on Zulip):

but all this can be discussed at a later occasion nbd

lqd (Oct 15 2019 at 19:11, on Zulip):

I also wondered about the plan about move/initialization errors, which is also the focus of Albin's work (and has opened a WIP PR)

lqd (Oct 15 2019 at 19:12, on Zulip):

how to architect it wrt rustc, especially seeing its impacts on the rest of the analysis

lqd (Oct 15 2019 at 19:13, on Zulip):

eg initialization and liveness greatly dominated the other analyses in a lot of examples I've tried

lqd (Oct 15 2019 at 19:13, on Zulip):

as mentioned in another thread I do have a way to make those 20-30x faster, but that would prevent computing all move errors for example, so depending on the plan, that's not going to be acceptable

lqd (Oct 15 2019 at 19:16, on Zulip):

so this week I'll probably look at the couple cases about the equality prototype and see where are the errors missed

nikomatsakis (Oct 15 2019 at 19:18, on Zulip):

argh sorry

lqd (Oct 15 2019 at 19:18, on Zulip):

maybe look at parsing, eg adding some of the missing relations or trying to figure something out about lalrpop, just to refresh my mind on them

nikomatsakis (Oct 15 2019 at 19:18, on Zulip):

man there's a lot to catch up here!

nikomatsakis (Oct 15 2019 at 19:18, on Zulip):

I probably won't be able to do that today

nikomatsakis (Oct 15 2019 at 19:18, on Zulip):

it's been a crazy day

lqd (Oct 15 2019 at 19:18, on Zulip):

not really don't worry about it :)

lqd (Oct 15 2019 at 19:19, on Zulip):

still about ! ?

lqd (Oct 15 2019 at 19:21, on Zulip):

if you find a couple minutes, Albin had questions about initialization IIUC, and since it might involve their thesis, it's probably more important than the rest

lqd (Oct 15 2019 at 19:22, on Zulip):

(I was also thinking about looking at illegal subset relations again, especially the placeholder loans strat, so we'll see if I can :)

lqd (Oct 15 2019 at 19:22, on Zulip):

if anyone has something to mention

lqd (Oct 15 2019 at 19:23, on Zulip):

otherwise good evening/afternoon everyone, ttyl :wave:

lqd (Oct 15 2019 at 19:23, on Zulip):

(thanks again for the answer @Matthew Jasper I'll look into it more)

Albin Stjerna (Oct 16 2019 at 06:51, on Zulip):

if you find a couple minutes, Albin had questions about initialization IIUC, and since it might involve their thesis, it's probably more important than the rest

Nope! I mean, it would be nice if I could add move errors to the thesis, but that's not my plan. :)

Albin Stjerna (Oct 16 2019 at 07:11, on Zulip):

maybe look at parsing, eg adding some of the missing relations or trying to figure something out about lalrpop, just to refresh my mind on them

I left a...lot of things to be done in polonius-parser, given that it's basically pre-liveness/initialisation. This is also scary because it makes testing my stuff harder.

Last update: Nov 15 2019 at 20:20UTC