Stream: t-compiler/wg-nll

Topic: pre-triage 2019.03.06


pnkfelix (Mar 06 2019 at 15:28, on Zulip):

I'll be doing some pre-triage for NLL in advance of tonight's weekly meeting

pnkfelix (Mar 06 2019 at 15:30, on Zulip):

as usual i'm following the structure presented in our NLL triage paper

pnkfelix (Mar 06 2019 at 15:32, on Zulip):

first, uncategorized issues

pnkfelix (Mar 06 2019 at 15:32, on Zulip):

only one item: "NLL: turn off migration mode" #58781

pnkfelix (Mar 06 2019 at 15:32, on Zulip):

(I added this as a subitem of "NLL: complete transition from migrate to full NLL" #57895)

pnkfelix (Mar 06 2019 at 15:33, on Zulip):

needs to be tagged as a tracking issue I think. I'll do that now.

pnkfelix (Mar 06 2019 at 15:33, on Zulip):

that takes it out of the "uncategorized issues" search. great.

pnkfelix (Mar 06 2019 at 15:34, on Zulip):

next, unprioritized issues; these need a prioritization (i.e. a P-label, like P-high)

pnkfelix (Mar 06 2019 at 15:36, on Zulip):

lets see. I'll try to attack these using the NLL-* labels as a guide for initial priorization.

pnkfelix (Mar 06 2019 at 15:37, on Zulip):

There's only one NLL-sound issue here: "fix "bivariant wf" bug in the NLL subtyping code" #54105

pnkfelix (Mar 06 2019 at 15:38, on Zulip):

I'm going to tag this as P-medium; until we actually see an instance of this bug (either by constructing one or in the wild), I don't want to keep revisiting it week after week.

pnkfelix (Mar 06 2019 at 15:39, on Zulip):

(However, it is possible that the @WG-compiler-traits group would disagree with that prioritization of #54105 ?)

pnkfelix (Mar 06 2019 at 15:40, on Zulip):

next, lets prioritize the NLL-complete issues

pnkfelix (Mar 06 2019 at 15:40, on Zulip):

first: "DerefMut borrow method call is too long with Deref arguments" #57376

pnkfelix (Mar 06 2019 at 15:41, on Zulip):

this is an issue with two-phased borrows

pnkfelix (Mar 06 2019 at 15:43, on Zulip):

I'm going to tag this as P-medium, same as "Tracking issue for generalized two-phase borrows" #49434

pnkfelix (Mar 06 2019 at 15:49, on Zulip):

next: "ICE when returning generic abitrary self type in defining function for existential associated type" #57700

pnkfelix (Mar 06 2019 at 15:50, on Zulip):

based on skimming the comments, this issue is believed to be the same as issue #53598

pnkfelix (Mar 06 2019 at 15:53, on Zulip):

I suspect the dependence on #![feature(existential _type)] should not be used as a justification for assigning a low priority.

pnkfelix (Mar 06 2019 at 15:55, on Zulip):

I've been having some amount of luck lately by changing bug!/span_bug! calls into tcx.sess.delay_span_bug to get rid of ICE's in cases where the compiler would issue a proper error if it hadn't encountered the ICE (which, I think, this issue falls into)

pnkfelix (Mar 06 2019 at 15:55, on Zulip):

I'll tag it as P-high and assign it to myself for initial investigation.

pnkfelix (Mar 06 2019 at 15:57, on Zulip):

okay that's all the NLL-complete issues

pnkfelix (Mar 06 2019 at 15:58, on Zulip):

there are six unprioritized NLL-diagnostics issues, and two NLL-performant ones

pnkfelix (Mar 06 2019 at 15:59, on Zulip):

I don't think I'm in a position right now to fairly evaluate priority of diagnostic issues, so I'll table those (again)

pnkfelix (Mar 06 2019 at 15:59, on Zulip):

but lets see if we can tie off the NLL-performant ones

pnkfelix (Mar 06 2019 at 15:59, on Zulip):

first: " [nll] hash borrows in scope for better performance" #53159

pnkfelix (Mar 06 2019 at 16:00, on Zulip):

During triage two weeks ago I stated I didn't know what priority to assign here, and niko said they were inclined to just close the issue.

pnkfelix (Mar 06 2019 at 16:00, on Zulip):

(its possible the idea is actionable/relevant, but I'm not sure that warrants keeping an issue open.)

pnkfelix (Mar 06 2019 at 16:01, on Zulip):

let me see if there's a comment in the code about it.

pnkfelix (Mar 06 2019 at 16:02, on Zulip):

the issue number does not occur in the source code. Its probably not worth even adding a comment about it.

pnkfelix (Mar 06 2019 at 16:02, on Zulip):

I'll just close.

pnkfelix (Mar 06 2019 at 16:03, on Zulip):

second (and last) NLL performant issue: "NLL compile-time performance regression" #58178

pnkfelix (Mar 06 2019 at 16:05, on Zulip):

The relevant benchmark has been added, or at least rustc-perf#343 has landed.

pnkfelix (Mar 06 2019 at 16:06, on Zulip):

ah but the perf web site says that its data is

pnkfelix (Mar 06 2019 at 16:06, on Zulip):

last updated on March 4th?

pnkfelix (Mar 06 2019 at 16:07, on Zulip):

which may well have predated the PR landing

pnkfelix (Mar 06 2019 at 16:07, on Zulip):

anyway it sounds like we likely still have a problem here.

pnkfelix (Mar 06 2019 at 16:08, on Zulip):

I'm going to tag this as P-high

pnkfelix (Mar 06 2019 at 16:10, on Zulip):

before I look over the P-high issues, I'll just note: "user type annotations are captured post normalization" #54940 is still tagged as I-nominated.

pnkfelix (Mar 06 2019 at 16:10, on Zulip):

So my Q for @nikomatsakis : can you or someone else provide a short statement (not necessarily today) as to the plan for lazy normalization?

pnkfelix (Mar 06 2019 at 16:11, on Zulip):

oh. There's only one unassigned P-high issue, namely, the one I just tagged as P-high: "NLL compile-time performance regression" #58178

pnkfelix (Mar 06 2019 at 16:12, on Zulip):

I'll nominate that for discussion at meeting tonight, to see if anyone wants to pick up where @Matthew Jasper has left off in terms of investigating the performance.

pnkfelix (Mar 06 2019 at 16:13, on Zulip):

and that's all of the pre-triage. See you all at the meeting tonight!

csmoe (Mar 06 2019 at 16:55, on Zulip):

@pnkfelix I'd like to take care of #58178 regression, but it's really late(1:00am) now in my timezone, so cannot attend later, would you mind pinging me at the beginning of the discussion of this issue in the upcoming meeting? (for me to quick-check what was going on tomorrow morning).
But if you find another candidate of this in the meeting, that's fine.

lqd (Mar 06 2019 at 18:32, on Zulip):

it would be cool if we could also talk tonight about the WG's permissions on the repo :)

pnkfelix (Mar 06 2019 at 19:30, on Zulip):

@csmoe I'll just go ahead and assign #58178 to you

Last update: Nov 22 2019 at 01:00UTC