Stream: t-compiler/wg-nll

Topic: removing-locations-all


nikomatsakis (May 23 2018 at 17:54, on Zulip):

So I have this PR that I want to land:

https://github.com/rust-lang/rust/pull/50938

However, it currently generates an ICE. I could use some help to dig into it, maybe?

nikomatsakis (May 23 2018 at 17:55, on Zulip):

Though actually as a very first step, we could just run with this PR to regenerate facts

nikomatsakis (May 23 2018 at 17:55, on Zulip):

and feed those into polonius

nikomatsakis (May 23 2018 at 17:57, on Zulip):

actually @Santiago Pastorino asking for help to diagnose this ICE might be cruel and unusual punishment

nikomatsakis (May 23 2018 at 17:57, on Zulip):

I think I sorta see the problem

nikomatsakis (May 23 2018 at 17:57, on Zulip):

looking again with fresh eyes

nikomatsakis (May 23 2018 at 17:57, on Zulip):

not 100% sure how to fix yet though

nikomatsakis (May 23 2018 at 17:57, on Zulip):

I should probably re-run to generate the issue-42780 and clap facts

Santiago Pastorino (May 23 2018 at 18:10, on Zulip):

ok

Santiago Pastorino (May 23 2018 at 18:10, on Zulip):

I'm all ears anyway :)

nikomatsakis (May 23 2018 at 18:13, on Zulip):

ok :) I'd say for now if you have time and feel like hacking the diagnostics stuff is good ... I'll figure this damn thing out ...

Santiago Pastorino (May 23 2018 at 18:15, on Zulip):

ok

Santiago Pastorino (May 23 2018 at 18:15, on Zulip):

I'm going to focus on diagnostics stuff and then ping the rest about rustc and polonius integration

nikomatsakis (May 24 2018 at 00:40, on Zulip):

curious. I regenerated the facts using that branch — which means far fewer outlives facts, among other things — and found that overall the analysis ran slightly slower (e.g., 7.2s instead of 6.5s for DatafromOpt on clap). I have no idea why that would be.

nikomatsakis (May 24 2018 at 00:41, on Zulip):
> git show --stat 65d091795b
commit 65d091795b43467a1a789ecca259c120c42528de
Author: Niko Matsakis <niko@alum.mit.edu>
Date:   Wed May 23 15:32:56 2018 -0400

    regenerate facts without `Location::All`

 .../clap-rs/app-parser-{{impl}}-add_defaults/invalidates.facts  |  62166 ----
 inputs/clap-rs/app-parser-{{impl}}-add_defaults/outlives.facts  | 490010 -------------------------
 .../app-parser-{{impl}}-add_defaults/region_live_at.facts       | 294012 +++++++++++++++
 inputs/issue-47680/nll-facts/main/region_live_at.facts          |    116 +
 inputs/issue-47680/nll-facts/main/universal_region.facts        |      2 -
 inputs/issue-47680/nll-facts/{{impl}}-maybe_next/outlives.facts |     12 -
 .../nll-facts/{{impl}}-maybe_next/region_live_at.facts          |     14 +
 .../nll-facts/{{impl}}-maybe_next/universal_region.facts        |      3 -
nikomatsakis (May 24 2018 at 00:42, on Zulip):

but the region_live_at facts were being added as a pre-processing step anyway...

nikomatsakis (May 24 2018 at 00:42, on Zulip):

(I also don't know why the set of invalidates facts changed)

Last update: Nov 22 2019 at 00:00UTC