Stream: t-compiler

Topic: pre-meeting triage 2020-03-05 #54818


pnkfelix (Mar 04 2020 at 17:24, on Zulip):

I will be doing pre-triage in this channel.

pnkfelix (Mar 04 2020 at 17:35, on Zulip):

first up, nominated unprioritized issues

pnkfelix (Mar 04 2020 at 17:35, on Zulip):

there are three such issues

pnkfelix (Mar 04 2020 at 17:36, on Zulip):

unpri nom 1/3: "Regression in rustc nightly" #69629

pnkfelix (Mar 04 2020 at 17:39, on Zulip):

Seems very bad

pnkfelix (Mar 04 2020 at 17:41, on Zulip):

triage: P-high, removing nomination. Assigning to @Esteban Küber initially (but feel free to clear assignment or reassign if you are overloaded)

pnkfelix (Mar 04 2020 at 17:45, on Zulip):

by the way, under the new priority scheme being considered for next generation triage, I would consider #69629 a P-critical bug. The regressing PR took code that used to compile successfully, and now produced an unexpected static error.

pnkfelix (Mar 04 2020 at 17:46, on Zulip):

unpri nom 2/3: "internal compiler error: Encountered ambiguity selecting Binder(<[type error] as TraitA>) during codegen, presuming due to overflow" #69602

pnkfelix (Mar 04 2020 at 17:52, on Zulip):

triage: already has fix in pipeline. P-high. removing nomination label. Assigning to @Esteban Küber as well.

pnkfelix (Mar 04 2020 at 17:53, on Zulip):

unpri nom 3/3: "can not compile in release mode in raspberry pi zero" #69420

pnkfelix (Mar 04 2020 at 18:00, on Zulip):

seems pretty bad to have such a simple test cause an LLVM failure in release mode.

pnkfelix (Mar 04 2020 at 18:01, on Zulip):

in this case, I'm going triage as P-high, and also leave it nominated in hopes of discussing at T-compiler meeting.

pnkfelix (Mar 04 2020 at 18:02, on Zulip):

(the intention of leaving it nominated is to establish how to prioritize this, since arm is tier 2 target. Is this really P-high? P-critical (under future priority scheme)? Or in fact P-medium, due to tier 2 status...?)

pnkfelix (Mar 04 2020 at 18:19, on Zulip):

next up: unprioritized beta regressions ... there are zero.

pnkfelix (Mar 04 2020 at 18:20, on Zulip):

next up: unprioritized nightly regressions ... there are two

pnkfelix (Mar 04 2020 at 18:20, on Zulip):

unpri ν regr 1/2: "Regression in error message quality for macro_rules involving $:ident" #69604

pnkfelix (Mar 04 2020 at 18:24, on Zulip):

triage: needs-bisect. Prioritizing as P-medium, based on assumption that the quality of diagnostics here can regress slightly and then get fixed in a future release if need be.

pnkfelix (Mar 04 2020 at 18:29, on Zulip):

unpri ν regr 2/2: "is_x86_feature_detected!("avx512f") fails to build on nightly" #68905

pnkfelix (Mar 04 2020 at 18:30, on Zulip):

ah right, this is not tagged with T-compiler

pnkfelix (Mar 04 2020 at 18:35, on Zulip):

(so I'm not going to care about trying to prioritize)

pnkfelix (Mar 04 2020 at 20:18, on Zulip):

((actually I ended up tagging as P-high, and wrote a note explaining why...))

pnkfelix (Mar 04 2020 at 20:39, on Zulip):

(I am now going through the I-nominated issues and selecting which ones to add to the hackmd for tomorrow's meeting.)

pnkfelix (Mar 04 2020 at 20:40, on Zulip):

#69307 is just nominated for AsyncAwait, so it does not belong on our list for the meeting.

pnkfelix (Mar 04 2020 at 20:44, on Zulip):

#68729 is something I nominated

pnkfelix (Mar 04 2020 at 20:45, on Zulip):

as described here:

Leaving nominated because I want to at least discuss 1. whether we need/want to collect data on instances of such breakage, and 2. whether we were too aggressive in deciding that going straight to an error without some mitigating mechanism was too aggressive.

pnkfelix (Mar 04 2020 at 20:46, on Zulip):

but I think I am going to just remove the nomination. this was not such a big deal in the end. The main question I would want resolved is whether we actually documented this breaking change somewhere visible (release notes from release blog post or something similar). But that does not need to be a discussion during the synchronous T-compiler meeting.

pnkfelix (Mar 04 2020 at 20:50, on Zulip):

I-nominated: " Box<dyn FnOnce> doesn't respect self alignment" #68304

pnkfelix (Mar 04 2020 at 21:02, on Zulip):

(adding to explicit nominations)

pnkfelix (Mar 04 2020 at 21:03, on Zulip):

I ended up adding #67705 (which is seeking help from a windows dev) and #16238 (which I am nominating to close but wanted to sanity check before doing so)

pnkfelix (Mar 05 2020 at 14:18, on Zulip):

okay, in the scant time before the meeting starts, I'll try to look over the P-high issues

pnkfelix (Mar 05 2020 at 14:20, on Zulip):

there are 57 open P-high issues, 32 of which are unassigned

pnkfelix (Mar 05 2020 at 14:21, on Zulip):

P-high 1/57: "Regression in rustc nightly" #69629

pnkfelix (Mar 05 2020 at 14:21, on Zulip):

fixed by PR #69717. Reopened by @lqd 5 hours ago ... but did PR #69717 not have a test ...?

pnkfelix (Mar 05 2020 at 14:22, on Zulip):

it did not have a test, indeed.

lqd (Mar 05 2020 at 14:22, on Zulip):

yes that’s why I reopened it

pnkfelix (Mar 05 2020 at 14:22, on Zulip):

yeah I was just surprised

pnkfelix (Mar 05 2020 at 14:22, on Zulip):

okay sounds good

lqd (Mar 05 2020 at 14:22, on Zulip):

it was closed by ”mistake”

lqd (Mar 05 2020 at 14:22, on Zulip):

(some difference between the commit message and the pr description)

pnkfelix (Mar 05 2020 at 14:23, on Zulip):

gotcha

pnkfelix (Mar 05 2020 at 14:23, on Zulip):

P-high 2/57: "internal compiler error: Encountered ambiguity selecting Binder(<[type error] as TraitA>) during codegen, presuming due to overflow" #69602

Santiago Pastorino (Mar 05 2020 at 14:24, on Zulip):

there's a fix in the pipeline

pnkfelix (Mar 05 2020 at 14:24, on Zulip):

has PR, (even inclludes test)

pnkfelix (Mar 05 2020 at 14:24, on Zulip):

yeah, great

pnkfelix (Mar 05 2020 at 14:24, on Zulip):

P-high 3/57: "mullvad-vpn build fails after jnix crate upgrade to v0.2.0" #69490

pnkfelix (Mar 05 2020 at 14:24, on Zulip):

has fix in PR #69591

pnkfelix (Mar 05 2020 at 14:25, on Zulip):

currently set for review by niko, who is pretty overloaded.

Santiago Pastorino (Mar 05 2020 at 14:25, on Zulip):

yep

pnkfelix (Mar 05 2020 at 14:25, on Zulip):

I might swing back around this PR later and see if I can review it myself

pnkfelix (Mar 05 2020 at 14:26, on Zulip):

P-high 4/57: "can not compile in release mode in raspberry pi zero" #69420

pnkfelix (Mar 05 2020 at 14:26, on Zulip):

something I should have done with this one yesterday is fire off a signal to the ICE-Breakers-LLVM

pnkfelix (Mar 05 2020 at 14:27, on Zulip):

argh I can never find the bot invocation for this

pnkfelix (Mar 05 2020 at 14:27, on Zulip):

found it: https://rust-lang.github.io/rustc-guide/ice-breaker/about.html#tagging-an-issue-for-an-ice-breaker-group

pnkfelix (Mar 05 2020 at 14:28, on Zulip):

okay pinged the grouop

pnkfelix (Mar 05 2020 at 14:28, on Zulip):

other than that, this is nominated for today's meeting. not much else we can do with it now

pnkfelix (Mar 05 2020 at 14:29, on Zulip):

P-high 5/57: "SIGSEGV when compiling regex_automata wasm32-unknown-emscripten" #69364

pnkfelix (Mar 05 2020 at 14:29, on Zulip):

I meant to look at this last week but I got distracted by fun with cargo-bisect-rustc

Santiago Pastorino (Mar 05 2020 at 14:31, on Zulip):

bisecting it would be nice :)

pnkfelix (Mar 05 2020 at 14:31, on Zulip):

it looks like it has gotten some reduction to a small test case, though it still depends on an external crate.

Santiago Pastorino (Mar 05 2020 at 14:31, on Zulip):

we can say you were preparing the tool to make this work :)

pnkfelix (Mar 05 2020 at 14:31, on Zulip):

I'm having a slightly hard time interpreting the latest comment; it sounds like very recent versions of the compiler do not reproduce the segfault

Santiago Pastorino (Mar 05 2020 at 14:32, on Zulip):

pnkfelix said:

it looks like it has gotten some reduction to a small test case, though it still depends on an external crate.

we can still bisect

Santiago Pastorino (Mar 05 2020 at 14:32, on Zulip):

yeah, I'm not sure what does it meant exactly neither

pnkfelix (Mar 05 2020 at 14:32, on Zulip):

so okay, its possible this is something that has either been fixed, or it is going to come and go intermittently.

pnkfelix (Mar 05 2020 at 14:32, on Zulip):

I'll leave it assigned to me and P-high

Santiago Pastorino (Mar 05 2020 at 14:32, on Zulip):

it's something that regressed after feb 27? or after feb 27 got fixed?

pnkfelix (Mar 05 2020 at 14:32, on Zulip):

but the actual severity remains unknown

pnkfelix (Mar 05 2020 at 14:33, on Zulip):

it sounds like it broke a long time ago

pnkfelix (Mar 05 2020 at 14:33, on Zulip):

or rather, 1.40.0

pnkfelix (Mar 05 2020 at 14:34, on Zulip):

and technical issues with 1.39 make it difficult to recreate the actual bug conditions

pnkfelix (Mar 05 2020 at 14:34, on Zulip):

anyway we can move along

pnkfelix (Mar 05 2020 at 14:34, on Zulip):

P-high 6/57: "ICE: src/librustc/middle/region.rs:1037: Encountered greater count 28" #69307

pnkfelix (Mar 05 2020 at 14:35, on Zulip):

okay looks like the AsyncAwait team have triaged this and know about it. and it was successfully reduced to a MCVE

pnkfelix (Mar 05 2020 at 14:36, on Zulip):

its an async-await bug; its possible its been there forever. but it would be best to establish that explicitly.

pnkfelix (Mar 05 2020 at 14:36, on Zulip):

and again, I've left the nomination label in place for AsyncAwait team

pnkfelix (Mar 05 2020 at 14:37, on Zulip):

but its possible it could be removed

pnkfelix (Mar 05 2020 at 14:37, on Zulip):

lets move along

pnkfelix (Mar 05 2020 at 14:37, on Zulip):

P-high 7/57: "Compiler bug: type parameter T/#0 (T/0) out of range when substituting" #69296

Santiago Pastorino (Mar 05 2020 at 14:38, on Zulip):

there's a MVCE at least :)

Santiago Pastorino (Mar 05 2020 at 14:39, on Zulip):

uncertain if that's a regression or not

pnkfelix (Mar 05 2020 at 14:39, on Zulip):

yeah I'm looking on godbolt now

Santiago Pastorino (Mar 05 2020 at 14:39, on Zulip):

it crashes on current stable

Santiago Pastorino (Mar 05 2020 at 14:39, on Zulip):

tried on play

Santiago Pastorino (Mar 05 2020 at 14:39, on Zulip):

did this ever worked?

pnkfelix (Mar 05 2020 at 14:40, on Zulip):

no, AFAICT it has crashed for as long as this these were made const fn's.

Santiago Pastorino (Mar 05 2020 at 14:40, on Zulip):

yep, just checked the same

pnkfelix (Mar 05 2020 at 14:40, on Zulip):

which to me means this is not a critical bug

pnkfelix (Mar 05 2020 at 14:41, on Zulip):

would I downgrade it to P-medium? I think not; my intuition is to keep it at P-high

Santiago Pastorino (Mar 05 2020 at 14:41, on Zulip):

that means ... exactly I was going to say P-medium

pnkfelix (Mar 05 2020 at 14:41, on Zulip):

as in, "yeah, it would be nice to fix this before next release"

pnkfelix (Mar 05 2020 at 14:42, on Zulip):

Santiago Pastorino said:

that means ... exactly I was going to say P-medium

so you think we should explicitly downgrade to P-medium?

pnkfelix (Mar 05 2020 at 14:42, on Zulip):

anyway I'll leave a comment summarizing these findings

Santiago Pastorino (Mar 05 2020 at 14:43, on Zulip):

ohh I thought you were saying that, well under the new priority scheme this would not be P-critical

Santiago Pastorino (Mar 05 2020 at 14:43, on Zulip):

probably would be ok P-high in the new scheme

Santiago Pastorino (Mar 05 2020 at 14:43, on Zulip):

with this scheme, do we want this popping up again and again?

pnkfelix (Mar 05 2020 at 14:43, on Zulip):

right, definitely not P-critical

Santiago Pastorino (Mar 05 2020 at 14:44, on Zulip):

in the new scheme is more clear, under this one, unsure, so maybe let's leave in the way it is?

pnkfelix (Mar 05 2020 at 14:45, on Zulip):

yeah, the distinction between P-high and P-medium under new scheme is quite fuzzy

pnkfelix (Mar 05 2020 at 14:45, on Zulip):

the big reason I would make this P-high rather than P-medium, is that the compiler ICE's without any useful diagnostic for the user

pnkfelix (Mar 05 2020 at 14:46, on Zulip):

so the end user may not have any idea where things went wrong

Santiago Pastorino (Mar 05 2020 at 14:46, on Zulip):

I guess having P-critical, P-high and no tags for the rest would suffice given our current capacity

pnkfelix (Mar 05 2020 at 14:47, on Zulip):

well, I'd keep at least one of P-medium or P-low, because its useful to demarcate that something has been triaged

Santiago Pastorino (Mar 05 2020 at 14:47, on Zulip):

but also well P-medium and P-low would be ok to have as ... exactly that :)

pnkfelix (Mar 05 2020 at 14:47, on Zulip):

great. okay so lets move along

pnkfelix (Mar 05 2020 at 14:48, on Zulip):

P-high 8/57: "Significant performance regression on the encoding benchmark" #69197

Santiago Pastorino (Mar 05 2020 at 14:49, on Zulip):

from the comments they know what affect performance but uncertain why

pnkfelix (Mar 05 2020 at 14:50, on Zulip):

yeah. but I'm basically in agreement with @ecstatic-morse : we can just close this at this point

pnkfelix (Mar 05 2020 at 14:50, on Zulip):

buffering stderr sufficed (PR #69227)

Santiago Pastorino (Mar 05 2020 at 14:51, on Zulip):

right

pnkfelix (Mar 05 2020 at 14:51, on Zulip):

P-high 9/57: "beta regression: ICE on Tried to access field 0 of union Layout" #69191

pnkfelix (Mar 05 2020 at 14:52, on Zulip):

Looks like its been narrowed to PR #67501. Thanks @LeSeulArtichaut and @Wesley Wiser !

oli (Mar 05 2020 at 14:52, on Zulip):

I haven't found the time for that one yet

Santiago Pastorino (Mar 05 2020 at 14:53, on Zulip):

it has an MVCE and we know where it has regressed so that's good

pnkfelix (Mar 05 2020 at 14:53, on Zulip):

okay. maybe I'll see if I can fix it today

pnkfelix (Mar 05 2020 at 14:53, on Zulip):

since we don't have much time before the next release

LeSeulArtichaut (Mar 05 2020 at 14:54, on Zulip):

I've tried to dive into the code but I don't have enough experience for that

pnkfelix (Mar 05 2020 at 14:54, on Zulip):

(and this is a stable-to-beta regression)

pnkfelix (Mar 05 2020 at 14:57, on Zulip):

okay i'll do one more before meeting starts

pnkfelix (Mar 05 2020 at 14:57, on Zulip):

P-high 10/57: "beta regression: $crate may not be imported" #69190

pnkfelix (Mar 05 2020 at 15:00, on Zulip):

work item remains as stated

pnkfelix (Mar 05 2020 at 15:00, on Zulip):

(see about a 0.27.1 release.)

Last update: Jun 04 2020 at 17:50UTC