Stream: t-compiler

Topic: pre-meeting triage 2019-10-17 #54818


pnkfelix (Oct 17 2019 at 12:19, on Zulip):

I will be doing pre-triage in this channel.

pnkfelix (Oct 17 2019 at 12:22, on Zulip):

first up: un-prioritized nominated issues.

pnkfelix (Oct 17 2019 at 12:22, on Zulip):

there are 5 un-prioritized noms

pnkfelix (Oct 17 2019 at 12:23, on Zulip):

unpri 1/5: 'ICE: "Tried to access field 0 of union with 0 fields"' #65462

pnkfelix (Oct 17 2019 at 12:24, on Zulip):

oy it is so tempting to try to dive in an make an MCVE right now

pnkfelix (Oct 17 2019 at 12:24, on Zulip):

lets call this P-high and self-assign for now.

pnkfelix (Oct 17 2019 at 12:25, on Zulip):

unpri 2/5: "rustc panic in 1.40.0-nightly (c27f7568b 2019-10-13)" #65411

pnkfelix (Oct 17 2019 at 12:26, on Zulip):

the bug filer says this happened while compiling rayon

pnkfelix (Oct 17 2019 at 12:30, on Zulip):

marking as E-needs-mcve

pnkfelix (Oct 17 2019 at 12:31, on Zulip):

(rayon builds fine for me locally at least ....)

pnkfelix (Oct 17 2019 at 12:32, on Zulip):

but if manually run -Z save-analysis, I see the same ICE

pnkfelix (Oct 17 2019 at 12:33, on Zulip):

lets call this P-high and tentatively assign to @Igor Matuszewski

pnkfelix (Oct 17 2019 at 12:35, on Zulip):

unpri 3/5: "Nightly ICE with non-const macro in const" #65394

pnkfelix (Oct 17 2019 at 12:35, on Zulip):

looks like @ecstatic-morse is on top of this one

pnkfelix (Oct 17 2019 at 12:36, on Zulip):

but is there going to be questions about what the semantics should be? See PR #65485

pnkfelix (Oct 17 2019 at 12:37, on Zulip):

anyway it looks to me like you should always get errors with diagnostics before you hit the ICE, if I understand issue #65394 correctly. So I'm calling this P-medium. (And assigning to @ecstatic-morse )

pnkfelix (Oct 17 2019 at 12:38, on Zulip):

unpri 4/5: "rust-lld since 1.38 overlaps .text with .rodata for embedded arm target" #65391

pnkfelix (Oct 17 2019 at 12:40, on Zulip):

this seems like it should be delegated to a specific subteam

pnkfelix (Oct 17 2019 at 12:40, on Zulip):

I'll post a note in the main meeting area and leave this nominated.

pnkfelix (Oct 17 2019 at 12:42, on Zulip):

unpri 5/5: "Could not find enclosing breakable with id HirId" #65383

pnkfelix (Oct 17 2019 at 12:43, on Zulip):

this produces a reasonable error before ICE'ing.

pnkfelix (Oct 17 2019 at 12:43, on Zulip):

I'm going to call it P-medium and assign to self.

centril (Oct 17 2019 at 12:44, on Zulip):

@Esteban Küber maybe should look at that... it feels like it could be due to diagnostics

pnkfelix (Oct 17 2019 at 12:44, on Zulip):

that's a good theory

pnkfelix (Oct 17 2019 at 12:45, on Zulip):

added @Esteban Küber to the assignee list too

pnkfelix (Oct 17 2019 at 12:45, on Zulip):

okay that's all the unprioritized nominated issues.

pnkfelix (Oct 17 2019 at 12:46, on Zulip):

we have zero unprioritized beta-regressions

pnkfelix (Oct 17 2019 at 12:46, on Zulip):

we have zero unprioritized nightly regressions

pnkfelix (Oct 17 2019 at 12:48, on Zulip):

now we do a quick pass over all the nominated issues to make sure that there are not any that should not be nominated (e.g. issues that should have been untagged at last week's meeting)

pnkfelix (Oct 17 2019 at 12:49, on Zulip):

#65401 and #65391 look appropriately nominated to me

pnkfelix (Oct 17 2019 at 12:50, on Zulip):

#65401 already has a PR (#65470), and it seems like the main thing blocking the PR now is that the author wants to know how to write a test that will reliably ICE going forward. :laughing:

pnkfelix (Oct 17 2019 at 12:50, on Zulip):

time to add an intrinsic, me thinks!

centril (Oct 17 2019 at 12:51, on Zulip):

oh noes... :fear:

pnkfelix (Oct 17 2019 at 12:51, on Zulip):

(I think they really want a reliable way to delay_span_bug, but either way, its funny)

pnkfelix (Oct 17 2019 at 12:52, on Zulip):

we do have #[rustc_error] already, so I could imagine variants on that to do what the author is asking for here..

pnkfelix (Oct 17 2019 at 12:54, on Zulip):

anyway if nothing else we can discuss that during the nomination portion of the main meeting.

pnkfelix (Oct 17 2019 at 12:54, on Zulip):

two left

pnkfelix (Oct 17 2019 at 12:54, on Zulip):

nom 3/4: "Star Import Precedence Regression in Beta" #65090

centril (Oct 17 2019 at 12:54, on Zulip):

(probably best to just discuss on PR)

pnkfelix (Oct 17 2019 at 12:55, on Zulip):

(probably best to just discuss on PR)

I think we can at least get a quick temperature reading of the room during the meeting. Doesn't need to be a design discussion.

pnkfelix (Oct 17 2019 at 12:55, on Zulip):

ah okay this (#65090) was something I nominated last week to double-check the priority assignment

pnkfelix (Oct 17 2019 at 12:56, on Zulip):

relevant log from last week

pnkfelix (Oct 17 2019 at 12:57, on Zulip):

no one responded. So I'm going to assume everyone who was present or read the meeting conversation after ward is okay with leaving it P-medium

pnkfelix (Oct 17 2019 at 12:57, on Zulip):

so I'm going to remove the nomination label.

centril (Oct 17 2019 at 12:58, on Zulip):

is this a language issue?

pnkfelix (Oct 17 2019 at 12:58, on Zulip):

is #56593 a language issue?

centril (Oct 17 2019 at 12:59, on Zulip):

no #65090

pnkfelix (Oct 17 2019 at 12:59, on Zulip):

sorry I was not clear

pnkfelix (Oct 17 2019 at 12:59, on Zulip):

@Vadim Petrochenkov asserted that #56593 should be a good representative for the long-standing bug that was reported in #65090

pnkfelix (Oct 17 2019 at 12:59, on Zulip):

so that's why I responded to your question with another question.

pnkfelix (Oct 17 2019 at 12:59, on Zulip):

(but I know my phrasing was ambiguous)

centril (Oct 17 2019 at 13:00, on Zulip):

ah -- it sounds like Vadim is saying that there is a design decision to be made

centril (Oct 17 2019 at 13:00, on Zulip):

and that #56593 is unfixable?

pnkfelix (Oct 17 2019 at 13:00, on Zulip):

"fixable" via "implementing globs properly" is my interpretation

centril (Oct 17 2019 at 13:02, on Zulip):

ah yes, they say so indeed

centril (Oct 17 2019 at 13:02, on Zulip):

sounds like a big project :P

pnkfelix (Oct 17 2019 at 13:02, on Zulip):

unfortunately the workaround suggested by @Vadim Petrochenkov on #65090 seems subtle, at least to me

pnkfelix (Oct 17 2019 at 13:03, on Zulip):

anyway, I'm removing nomination tag from #65090

pnkfelix (Oct 17 2019 at 13:03, on Zulip):

nom 4/4: "linking of libtest failed" #64872

pnkfelix (Oct 17 2019 at 13:03, on Zulip):

I nominated this today, so I assert it should remain nominated for the meeting itself. :)

pnkfelix (Oct 17 2019 at 13:04, on Zulip):

okay, T-minus 56 minutes, and we're at the P-high traversal

pnkfelix (Oct 17 2019 at 13:05, on Zulip):

but this week there are "only" 25 open P-high issues, so maybe I have a chance of getting through them

pnkfelix (Oct 17 2019 at 13:05, on Zulip):

P-high 1/25: " ICE resolving non-existent PartialEq::Eq from match of const" #65466

pnkfelix (Oct 17 2019 at 13:06, on Zulip):

I filed this largely as a procedural matter so that i could distinguish this case from #61188 (and I then closed the latter). I'm making progress here; my weekend crisis meant I didn't get my PR up, but its coming.

pnkfelix (Oct 17 2019 at 13:06, on Zulip):

P-high 2/25: 'ICE: "Tried to access field 0 of union with 0 fields"' #65462

pnkfelix (Oct 17 2019 at 13:07, on Zulip):

self-assigned today. Anyone watching who wants to work-steal, be my guest!

pnkfelix (Oct 17 2019 at 13:07, on Zulip):

(I don't expect to look at it until tomorrow at the earliest.)

pnkfelix (Oct 17 2019 at 13:07, on Zulip):

P-high 3/25: "rustc panic in 1.40.0-nightly (c27f7568b 2019-10-13)' #65411

pnkfelix (Oct 17 2019 at 13:08, on Zulip):

assigned to @Igor Matuszewski earlier today

pnkfelix (Oct 17 2019 at 13:08, on Zulip):

P-high 4/25: "Deferred ICEs (i.e. delay_span_bug) are not preserved by incremental." #65401

pnkfelix (Oct 17 2019 at 13:08, on Zulip):

already discussed as a nominated issue. We'll get to it later.

pnkfelix (Oct 17 2019 at 13:09, on Zulip):

same for P-high 5/25: #64872

centril (Oct 17 2019 at 13:09, on Zulip):

I filed this largely as a procedural matter so that i could distinguish this case from #61188 (and I then closed the latter). I'm making progress here; my weekend crisis meant I didn't get my PR up, but its coming.

We (the lang team) should probably look at stuctural_match and friends more broadly before we paint ourselves into a corner with a series of PRs

pnkfelix (Oct 17 2019 at 13:09, on Zulip):

We (the lang team) should probably look at stuctural_match and friends more broadly before we paint ourselves into a corner with a series of PRs

Yes I've been thinking the same

centril (Oct 17 2019 at 13:09, on Zulip):

(for now I think we should take a conservative stance with PRs to leave that design room open)

pnkfelix (Oct 17 2019 at 13:14, on Zulip):

(moved forked discussion to separate topic)

pnkfelix (Oct 17 2019 at 13:15, on Zulip):

P-high 6/25: "debuginfo/pretty-uninitialized-vec fails with Cannot access memory at address 0x7fffff7fe000" #64343

centril (Oct 17 2019 at 13:15, on Zulip):

would it be terrible if we just removed that one test? :D

centril (Oct 17 2019 at 13:15, on Zulip):

it is basically what I need to test rollups locally

pnkfelix (Oct 17 2019 at 13:16, on Zulip):

I don't know. Can we stick with the plan to talk about this tomorrow?

centril (Oct 17 2019 at 13:16, on Zulip):

Ok

pnkfelix (Oct 17 2019 at 13:16, on Zulip):

In fact, lets make this deal

pnkfelix (Oct 17 2019 at 13:16, on Zulip):

that I'll post on the issue

pnkfelix (Oct 17 2019 at 13:16, on Zulip):

if we don't talk about this issue at tomorrow's design meeting

pnkfelix (Oct 17 2019 at 13:17, on Zulip):

then that will imply that it is so unimportant that we should just remove it as you suggest

pnkfelix (Oct 17 2019 at 13:17, on Zulip):

(or mark it // ignore)

centril (Oct 17 2019 at 13:17, on Zulip):

:+1:

pnkfelix (Oct 17 2019 at 13:17, on Zulip):

This is like the nuclear option

pnkfelix (Oct 17 2019 at 13:17, on Zulip):

which has worked oh so well for e.g. Brexit

pnkfelix (Oct 17 2019 at 13:18, on Zulip):

okay next up

pnkfelix (Oct 17 2019 at 13:18, on Zulip):

P-high 7/25: "ThinLTO and -o compiler option can lead to duplicate object file inclusion in staticlib" #64153

pnkfelix (Oct 17 2019 at 13:18, on Zulip):

hey great looks like @mw found a fix for this (PR #65435)

Pietro Albini (Oct 17 2019 at 13:18, on Zulip):

then that will imply that it is so unimportant that we should just remove it as you suggest

https://twitter.com/Argorak/status/1093167206857404416

pnkfelix (Oct 17 2019 at 13:19, on Zulip):

"the perfect is the enemy of the good" ?

pnkfelix (Oct 17 2019 at 13:19, on Zulip):

/shrug

pnkfelix (Oct 17 2019 at 13:20, on Zulip):

P-high 8/25: "STATUS_ACCESS_VIOLATION and STATUS_HEAP_CORRUPTION during compilation" #63959

pnkfelix (Oct 17 2019 at 13:21, on Zulip):

hmm. the latest comment from mati865 makes me wonder if this is a CI tooling issue?

pnkfelix (Oct 17 2019 at 13:21, on Zulip):

there was some other discussion elsewhere about upgrading our version of GCC on the CI

pnkfelix (Oct 17 2019 at 13:21, on Zulip):

right, PR #65302

pnkfelix (Oct 17 2019 at 13:22, on Zulip):

I wonder if there's any chance that might resolve this ... ?

pnkfelix (Oct 17 2019 at 13:22, on Zulip):

but no, the issue here in #63959 is using MSVC

pnkfelix (Oct 17 2019 at 13:23, on Zulip):

but then again the issue filer says it happens on both MSVC and GNU builds ... hmm...

pnkfelix (Oct 17 2019 at 13:24, on Zulip):

well anyway it seems like we're sort of stuck here, like last week

pnkfelix (Oct 17 2019 at 13:25, on Zulip):

I'll ask @Mateusz Mikuła if they could potentially experiment with CI docker images, maybe ...

pnkfelix (Oct 17 2019 at 13:26, on Zulip):

P-high 9/25: "Associated types, impl traits ~and closures~; oh my, an ICE." #63154

pnkfelix (Oct 17 2019 at 13:27, on Zulip):

I had one PR that "fixed" this, then @eddyb identified a variant that still ICE's and reopened it.

pnkfelix (Oct 17 2019 at 13:28, on Zulip):

which reminds me that I wanted to float my policy question about whether we, as a team, are better off reopening bugs in such cases, or if we should strive to open fresh issues (and link to the closed related one)

Mateusz Mikuła (Oct 17 2019 at 13:28, on Zulip):

I'll ask Mateusz Mikuła if they could potentially experiment with CI docker images, maybe ...

Windows builds do not use docker.

pnkfelix (Oct 17 2019 at 13:29, on Zulip):

ah, shows what I know

pnkfelix (Oct 17 2019 at 13:32, on Zulip):

still, it seems like the problem in #63959 must be due to some environment difference between the two build environments, right?

simulacrum (Oct 17 2019 at 13:33, on Zulip):

It should definitely be possible to get a very similar environment (e.g., by doing exactly what we do on CI on Windows)

pnkfelix (Oct 17 2019 at 13:34, on Zulip):

anyway I guess I need to decide if I'm going to leave #63154 assigned to me

pnkfelix (Oct 17 2019 at 13:34, on Zulip):

or if I'm going to try to find someone else to try to fix it

pnkfelix (Oct 17 2019 at 13:36, on Zulip):

P-high 10/25: " incremental compilation DepNode Bug running cargo check" #63150

pnkfelix (Oct 17 2019 at 13:37, on Zulip):

I continue to think its going to be nigh impossible to act on #63150 without more info

pnkfelix (Oct 17 2019 at 13:37, on Zulip):

but I hold out hope that it is a duplicate of #62649, which does have enough information for me to reproduce it.

pnkfelix (Oct 17 2019 at 13:37, on Zulip):

P-high 11/25: "Segfault compiling libc on armv7-unknown-linux-gnueabihf" #62896

pnkfelix (Oct 17 2019 at 13:39, on Zulip):

no real progress here yet

pnkfelix (Oct 17 2019 at 13:39, on Zulip):

(but I did get an account approved on the gcc compile farm this week, so I might be assist with these things in the near future...)

pnkfelix (Oct 17 2019 at 13:40, on Zulip):

P-high 12/25: "Undefined symbol _fltused when compiling to x86_64-unknown-uefi" #62785

pnkfelix (Oct 17 2019 at 13:40, on Zulip):

hmm https://github.com/rust-lang/compiler-builtins/pull/317 still has not landed

pnkfelix (Oct 17 2019 at 13:41, on Zulip):

@oli I infer that @Alex Crichton thinks the PR needs some more work to properly hook up to the CI ?

pnkfelix (Oct 17 2019 at 13:41, on Zulip):

P-high 12/25: "ICE: Generic type alias to invalid type crashes during type check on latest stable" #62742

oli (Oct 17 2019 at 13:41, on Zulip):

yes, I've been busy and didn't get to any Rust stuff in the last 7 days

pnkfelix (Oct 17 2019 at 13:41, on Zulip):

@oli I'm in a similar boat

pnkfelix (Oct 17 2019 at 13:42, on Zulip):

regarding #62742, @nikomatsakis posted a big ol' comment 8 days ago ...

pnkfelix (Oct 17 2019 at 13:43, on Zulip):

anyway it seems like its progressing

pnkfelix (Oct 17 2019 at 13:44, on Zulip):

P-high 13/25: "internal compiler error: src/librustc/dep_graph/graph.rs:688: DepNode Hir(...) should have been pre-allocated but wasn't." #62649

pnkfelix (Oct 17 2019 at 13:44, on Zulip):

I started looking at this today.

pnkfelix (Oct 17 2019 at 13:44, on Zulip):

hopefully will be able to allocate more time to it either tonight or tomorrow

pnkfelix (Oct 17 2019 at 13:44, on Zulip):

P-high 14/25: "couldn't load codegen backend on windows-gnu" #61561

pnkfelix (Oct 17 2019 at 13:45, on Zulip):

I think its time for me to be honest with my self: with my current work-load and development set up, I don't think its reasonable for me to keep this bug assigned to myself

pnkfelix (Oct 17 2019 at 13:45, on Zulip):

I'll unassign myself and post a work request in the main meeting topic.

pnkfelix (Oct 17 2019 at 13:48, on Zulip):

P-high 15/25: "1.30 -> 1.31 dylib late-binding regression with GNU binutils 2.28 or older." #61539

pnkfelix (Oct 17 2019 at 13:48, on Zulip):

no progress made here either

pnkfelix (Oct 17 2019 at 13:51, on Zulip):

@nagisa said last week that they didn't think there was anything actionable to do on this issue

pnkfelix (Oct 17 2019 at 13:51, on Zulip):

but it just seems so ... bad

pnkfelix (Oct 17 2019 at 13:53, on Zulip):

even just sniffing the linker version when you ask for --crate-type=dylib seems like it would be a step forward ...

pnkfelix (Oct 17 2019 at 13:54, on Zulip):

P-high 16/25: "ICE librustc_mir/monomorphize/collector.rs:776: Cannot create local mono-item for DefId" #60363

pnkfelix (Oct 17 2019 at 13:54, on Zulip):

I haven't had a chance to look at this

pnkfelix (Oct 17 2019 at 13:55, on Zulip):

P-high 17/25: "Incremental compilation results in linker error when method use is removed" #59535

pnkfelix (Oct 17 2019 at 13:55, on Zulip):

I haven't had a chance to look at this

pnkfelix (Oct 17 2019 at 13:57, on Zulip):

/me sits and ponders what policy should be about unassigning bugs whose owners admit no progress on. It would be better to have a clean set of issues that are "available for people to take"

pnkfelix (Oct 17 2019 at 13:57, on Zulip):

but I know personally I like to have the integrated list of issues that I said "I think i can handle that eventually..."

pnkfelix (Oct 17 2019 at 13:57, on Zulip):

P-high 18/25: " Coherence can be bypassed by an indirect impl for a trait object" #57893

pnkfelix (Oct 17 2019 at 13:58, on Zulip):

my understand is that work has continued to stall here; is that correct, @nikomatsakis ?

pnkfelix (Oct 17 2019 at 13:58, on Zulip):

P-high 19/25: "ICE from const item in lifetime-parametric impl" #56445

centril (Oct 17 2019 at 13:59, on Zulip):

no activity on coherence in the wg-traits stream atm

pnkfelix (Oct 17 2019 at 13:59, on Zulip):

huh. PlasmaPower posted an interesting comment

pnkfelix (Oct 17 2019 at 14:00, on Zulip):

P-high 20/25: "ICE in macro: doc meta with expr on an item, string concat, stringify!(...)" #55414

centril (Oct 17 2019 at 14:01, on Zulip):

looks like a lazy norm issue

pnkfelix (Oct 17 2019 at 14:01, on Zulip):

@Vadim Petrochenkov tagged this as E-mentor last week

pnkfelix (Oct 17 2019 at 14:01, on Zulip):

looks like a lazy norm issue

(I assume this was referring to #57893 )

centril (Oct 17 2019 at 14:02, on Zulip):

no #56445

pnkfelix (Oct 17 2019 at 14:02, on Zulip):

oh okay

centril (Oct 17 2019 at 14:02, on Zulip):

basically fn foo() -> [X; Self::PROJ];

pnkfelix (Oct 17 2019 at 14:03, on Zulip):

yeah I guess that's true

pnkfelix (Oct 17 2019 at 14:03, on Zulip):

might as well add the tag

pnkfelix (Oct 17 2019 at 14:03, on Zulip):

okay, well, that's 4/5's of the issues. <<sigh>>

pnkfelix (Oct 17 2019 at 14:04, on Zulip):

(had to decide between starting triage early, or actually making progress on any of the issues I had been assigned first...)

Last update: Nov 16 2019 at 01:15UTC