Stream: t-compiler

Topic: pre-meeting triage 2019-09-12 #54818


pnkfelix (Sep 12 2019 at 10:59, on Zulip):

I will be doing pre-triage in this channel.

pnkfelix (Sep 12 2019 at 11:04, on Zulip):

So lets see if I still have the muscle memory for this

pnkfelix (Sep 12 2019 at 11:04, on Zulip):

first up: unprioritized nominated issues

pnkfelix (Sep 12 2019 at 11:04, on Zulip):

unpri nom 1/3: "miri no longer builds after rust-lang/rust#64354" #64363

pnkfelix (Sep 12 2019 at 11:06, on Zulip):

tagging as P-medium and unnominating.

pnkfelix (Sep 12 2019 at 11:07, on Zulip):

unpri nom 2/3: "ThinLTO and -o compiler option can lead to duplicate object file inclusion in staticlib" #64153

pnkfelix (Sep 12 2019 at 11:09, on Zulip):

I am of mixed minds here

pnkfelix (Sep 12 2019 at 11:09, on Zulip):

in terms of how to prioritize this

pnkfelix (Sep 12 2019 at 11:09, on Zulip):

the main danger here is to clients using external build systems (i.e., not cargo)

pnkfelix (Sep 12 2019 at 11:11, on Zulip):

it also just seems like something fragile in general?

pnkfelix (Sep 12 2019 at 11:12, on Zulip):

I'm inclined to tag as P-high. @mw, do you have any thoughts? Do you have any thoughts as to how we should fix this?

pnkfelix (Sep 12 2019 at 11:13, on Zulip):

from the comments that @mw left on the ticket, it seems like they at least have some notion of how a good fix would behave (regardless of how that fix is implemented)

pnkfelix (Sep 12 2019 at 11:13, on Zulip):

so I'm going to tag this as P-high, and speculatively assign it to @mw, with the intention that they choose if they'd prefer to fix it themselves, or to mentor someone else in authoring a fix.

pnkfelix (Sep 12 2019 at 11:15, on Zulip):

unpri nom 3/3: "Regression: Matching on function pointers fails when one of the formal params implements custom PartialEq" #63479

pnkfelix (Sep 12 2019 at 11:16, on Zulip):

hmm apparently @nikomatsakis was hoping that I would look at this. :smiley:

pnkfelix (Sep 12 2019 at 11:17, on Zulip):

it was marked as a stable-to-beta regression four weeks ago

pnkfelix (Sep 12 2019 at 11:17, on Zulip):

which means we don't have much time to fix it if we want to land a fix and beta-backport it before the next beta->stable transition

pnkfelix (Sep 12 2019 at 11:18, on Zulip):

I'll mark as P-high and assign to self. Lets see if I can come up with something quickly.

pnkfelix (Sep 12 2019 at 11:22, on Zulip):

next up: unprioritized stable-to-nightly regressions

pnkfelix (Sep 12 2019 at 11:22, on Zulip):

there's just one unpri nightly regression: "Don't know the field number in this context" #64385

pnkfelix (Sep 12 2019 at 11:25, on Zulip):

I'll just tag this as P-high; its an ICE and we don't want it to leak to beta. Plus @Vadim Petrochenkov self-assigned and says it will have a fix soon.

pnkfelix (Sep 12 2019 at 11:26, on Zulip):

Okay, next, we have forty-one P-high issues

pnkfelix (Sep 12 2019 at 11:27, on Zulip):

we can skip #64385 and #64153 since we just discussed them.

pnkfelix (Sep 12 2019 at 11:27, on Zulip):

P-high 3/41: "redundant_semicolon and clippy::no_effect tripped without emitting line numbers #63967

pnkfelix (Sep 12 2019 at 11:28, on Zulip):

sweet @Nathan Whitaker has a PR up to fix this (#6438)

pnkfelix (Sep 12 2019 at 11:29, on Zulip):

P-high 4/41: "STATUS_ACCESS_VIOLATION and STATUS_HEAP_CORRUPTION during compilation" #63959

pnkfelix (Sep 12 2019 at 11:32, on Zulip):

hmm. From the comment thread it sounds like @Mateusz Mikuła (github: mati865) has been working hard to try to investigate this, but is current blocked even just building the compiler due to issue #61561

pnkfelix (Sep 12 2019 at 11:33, on Zulip):

no one has marked this as a regression, but people are saying that it seems to be related to an LLVM upgrade from back in July. So it probably is a regression and we just have failed to mark it as such?

pnkfelix (Sep 12 2019 at 11:34, on Zulip):

Meanwhile the blocking issue #61561 is not marked with any priority at all

pnkfelix (Sep 12 2019 at 11:35, on Zulip):

Lets fix at least that.

pnkfelix (Sep 12 2019 at 11:37, on Zulip):

(it seems like the issue there may be an incompatibility with gcc 9.x (versus 8.x) on windows mingw? Definitely would be nice to have someone figure out what's going on there, and determine if this is something we should fix, or just document as broken.

pnkfelix (Sep 12 2019 at 11:37, on Zulip):

anyway I'll move along at this point

pnkfelix (Sep 12 2019 at 11:37, on Zulip):

P-high 5/41: "[1.38] Regression: cannot find macro trace! in this scope" #63888

pnkfelix (Sep 12 2019 at 11:39, on Zulip):

closing as fixed due to backport landing PR #64097

pnkfelix (Sep 12 2019 at 11:39, on Zulip):

P-high 6/41: "rustc could not initialize thread_rng (on pre-getrandom Linux kernel)" #63848

pnkfelix (Sep 12 2019 at 11:40, on Zulip):

Okay supposedly this will be fixed once PR #64381 lands

pnkfelix (Sep 12 2019 at 11:41, on Zulip):

(thanks to @Mateusz Mikuła (github: mati865)!)

pnkfelix (Sep 12 2019 at 11:42, on Zulip):

P-high 7/41: "Weird compiler behavior with in-lining of a function -> constant value arbitrary changed" #63516

pnkfelix (Sep 12 2019 at 11:44, on Zulip):

I don't have time to read all the dialogue on this issue right now; it seems like there may be debate about whether this is actually a compiler bug or not? Can someone take on the task of reviewing the comments and answering that Q for me?

pnkfelix (Sep 12 2019 at 11:44, on Zulip):

I'm going to leave this tagged as I-nominated in the hopes that, if no one answers the above Q here, then maybe we can talk about it in the meeting.

pnkfelix (Sep 12 2019 at 11:45, on Zulip):

P-high 8/41: "Regression: Matching on function pointers fails when one of the formal params implements custom PartialEq" #63479; skipping since we visited that up above.

pnkfelix (Sep 12 2019 at 11:45, on Zulip):

P-high 9/41: "Associated types, impl traits and closures; oh my, an ICE." #63154

pnkfelix (Sep 12 2019 at 11:47, on Zulip):

most relevant comment is probably this one from @nagisa :

The fact that miri works and a regular compile does not suggests a missing call to normalize() somewhere, which would be fairly easy to fix once it is found.

pnkfelix (Sep 12 2019 at 11:48, on Zulip):

Normally I might consider just assigning something like this to myself, but I think the other bug I self-assigned up above is much higher priority than this

pnkfelix (Sep 12 2019 at 11:51, on Zulip):

anyway I'm not going to invest more time in it now, but it might be an interesting one for people who want to get their feet wet with the compiler (e.g. digging into where the missing normalization call is, assuming that is indeed the issue)

pnkfelix (Sep 12 2019 at 11:51, on Zulip):

(It would probably also be good just to get someone assigned to this. But as noted above, I'm hesistant to assign myself at the moment; better to focus on things impacting the current beta, IMO)

pnkfelix (Sep 12 2019 at 11:53, on Zulip):

P-high 10/41: "Bug running cargo check" #63150

pnkfelix (Sep 12 2019 at 11:53, on Zulip):

eek some sort of incremental dep node bug

pnkfelix (Sep 12 2019 at 11:54, on Zulip):

and its unassigned. @mw do you want me to assign this to you? I could take it and fold it into my more general investigation of incremental compilation issues.

pnkfelix (Sep 12 2019 at 11:55, on Zulip):

(i guess that's my way of saying: @mw , if you don't want it, then go ahead and assign it to me.)

pnkfelix (Sep 12 2019 at 11:55, on Zulip):

P-high 11/41: "Segfault compiling libc on nightly-armv7-unknown-linux-gnueabihf" #62896

pnkfelix (Sep 12 2019 at 11:57, on Zulip):

There's been some investigation and dialogue. @Mateusz Mikuła asked @msizanoen to try again with some changes to build config.

pnkfelix (Sep 12 2019 at 11:58, on Zulip):

it remains unassigned and I'm going to leave it that way for now, see if it progresses on its own

pnkfelix (Sep 12 2019 at 11:58, on Zulip):

though ... there is the one issue that this is also a stable-to-beta regression, which means we don't really want to wait another week for results if we can help it

pnkfelix (Sep 12 2019 at 11:59, on Zulip):

(at what point do we consider reverting the LLVM update...?)

pnkfelix (Sep 12 2019 at 11:59, on Zulip):

actually I will assign this to myself

pnkfelix (Sep 12 2019 at 12:02, on Zulip):

(if it really is an ARM-specific segfault, though, that may impede my ability to assist.)

mw (Sep 12 2019 at 12:04, on Zulip):

#63150 seems like a duplicate, let me take a look...

pnkfelix (Sep 12 2019 at 12:04, on Zulip):

P-high 12/41: "Undefined symbol _fltused when compiling to x86_64-unknown-uefi" #62785

pnkfelix (Sep 12 2019 at 12:04, on Zulip):

assigning to self to at least take care of bisection.

mw (Sep 12 2019 at 12:04, on Zulip):

https://github.com/rust-lang/rust/issues/62649 looks like the same as #63150

pnkfelix (Sep 12 2019 at 12:06, on Zulip):

okay. @Aaron Turon , if you have any progress to report regarding #62649, please post a comment on that issue.

pnkfelix (Sep 12 2019 at 12:07, on Zulip):

(I may independently attempt to ping aaron at some later point. We'll see; my evening schedule has become a bit more difficult of late.)

pnkfelix (Sep 12 2019 at 12:07, on Zulip):

P-high 13/41: "ICE: Generic type alias to invalid type crashes during type check on latest stable" #62742

pnkfelix (Sep 12 2019 at 12:08, on Zulip):

assigned to @nikomatsakis 28 days ago, but no further progress noted.

pnkfelix (Sep 12 2019 at 12:09, on Zulip):

It would be good if something could at least confirm if PR #48995 is indeed the culprit.

mw (Sep 12 2019 at 12:09, on Zulip):

@pnkfelix I won't be able to make the meeting today, so I'd like to bring #63742 to attention now

pnkfelix (Sep 12 2019 at 12:09, on Zulip):

but since its a stable-to-stable regression I'm not going to spendmore time on it now

pnkfelix (Sep 12 2019 at 12:10, on Zulip):

@mw hmm okay

mw (Sep 12 2019 at 12:10, on Zulip):

I don't actually know much about #63742, but it looks like it's pointing to a potentially bigger underlying issue

mw (Sep 12 2019 at 12:11, on Zulip):

that is: the compiler re-ordering code across generator yield points

pnkfelix (Sep 12 2019 at 12:11, on Zulip):

yeah okay

pnkfelix (Sep 12 2019 at 12:11, on Zulip):

I posted a message in the actual meeting channel to ensure we get to it

mw (Sep 12 2019 at 12:11, on Zulip):

it might be a t-lang issue, but maybe it's just a bug

mw (Sep 12 2019 at 12:11, on Zulip):

thanks!

pnkfelix (Sep 12 2019 at 12:11, on Zulip):

(since I know the triage channel is basically a wall-of-text)

pnkfelix (Sep 12 2019 at 12:12, on Zulip):

P-high 14/41: "internal compiler error: src/librustc/dep_graph/graph.rs:688: DepNode Hir(...) should have been pre-allocated but wasn't." #62649 -- we can "skip" this one since we just discussed it, or at least pinged someone about it.

pnkfelix (Sep 12 2019 at 12:13, on Zulip):

P-high 15/41: "Forgone caching in cycles caused much overflow in trait solving" #61960

pnkfelix (Sep 12 2019 at 12:13, on Zulip):

wow I'm amazed we haven't just given up on this issue and closed it.

pnkfelix (Sep 12 2019 at 12:13, on Zulip):

I mean, isn't that what you're supposed to do when I go on leave? Close all my pet issues? :wink:

pnkfelix (Sep 12 2019 at 12:14, on Zulip):

anyway I don't think anyone has anything to report. @nikomatsakis self-assigned it 28 days ago, but I suspect they have been busy with many other things.

pnkfelix (Sep 12 2019 at 12:15, on Zulip):

P-high 16/41: "Self as default type isnt typechecked" #61631

pnkfelix (Sep 12 2019 at 12:18, on Zulip):

Okay, so after reviewing the dialogue, it seems like there were repeatedly comments that claimed a broader issue here than just the handling of Self, but the lang team has reviewed those cases and come to the conclusion that those were expected behavior.

pnkfelix (Sep 12 2019 at 12:18, on Zulip):

it is just the Self bug that remains a bug

nikomatsakis (Sep 12 2019 at 12:18, on Zulip):

P-high 13/41: "ICE: Generic type alias to invalid type crashes during type check on latest stable" #62742

I keep meaning to up the priority of these bugs. Hard to juggle that.

pnkfelix (Sep 12 2019 at 12:19, on Zulip):

@Alexander Regueiro I see that you posted a comment here 5 days ago asking if anyone was tackling the issue. But I would have inferred, from your comment on July 11, that you, @Alexander Regueiro , were the most obvious candidate to tackle the issue (with just Self)

pnkfelix (Sep 12 2019 at 12:19, on Zulip):

I'm going to change the assignee list so that its assigned to @Alexander Regueiro and myself rather than @eddyb and myself.

eddyb (Sep 12 2019 at 12:21, on Zulip):

oh, the Self stuff

pnkfelix (Sep 12 2019 at 12:21, on Zulip):

P-high 17/41: "regression: rustdoc resolution changes" #61560

pnkfelix (Sep 12 2019 at 12:21, on Zulip):

has this become a stable-to-stable regression at this point? it was, long ago

pnkfelix (Sep 12 2019 at 12:21, on Zulip):

(surely we've had a beta->stable uplift since July 11 ...)

pnkfelix (Sep 12 2019 at 12:22, on Zulip):

anyway leaving assigned to nagisa. Also, wondering whether this is even P-high at this point

pnkfelix (Sep 12 2019 at 12:23, on Zulip):

P-high 18/41: "1.30 -> 1.31 dylib late-binding regression with GNU binutils 2.28 or older." #61539

pnkfelix (Sep 12 2019 at 12:24, on Zulip):

ugh, I guess this hasn't really been discussed in a productive manner since ~July 11th?

pnkfelix (Sep 12 2019 at 12:25, on Zulip):

well its still assigned to me and @nagisa for now

pnkfelix (Sep 12 2019 at 12:26, on Zulip):

I'm going to leave it P-high and I'm not going to nominate it yet. But my hope is that over the next week I'll review the situation again, and decide on some course of action to propose (that, or nominate it so that you all can discuss it at the meeting)

pnkfelix (Sep 12 2019 at 12:26, on Zulip):

P-high 19/41: "Creating a recursive type with infinite size leads to ICE" #61323

pnkfelix (Sep 12 2019 at 12:28, on Zulip):

normally I'd suggest downgrading bugs with handling of infinite-sized types to P-medium, but given that this is an incremental compilation (or at least dep-graph) bug, with a terrible error message, I continue to think this should be P-high.

pnkfelix (Sep 12 2019 at 12:30, on Zulip):

@nikomatsakis it might be good, when you remove nomination labels, to leave a comment saying the justification. In this case, I cannot tell if you removed the nomination because you intended to investigate, or because you intended to find someone else to do the investigation, or because you intended to deprioritize, etc...

pnkfelix (Sep 12 2019 at 12:31, on Zulip):

anyway, regarding #61323, I unassigned @nikomatsakis and assigned it to @Aaron Turon and myself.

nikomatsakis (Sep 12 2019 at 12:32, on Zulip):

yeah @pnkfelix good point, I do try to leave comments whenever replacing labels but...

pnkfelix (Sep 12 2019 at 12:32, on Zulip):

P-high 20/41: "ICE when trying to match on non-PartialEq slice." #61188

pnkfelix (Sep 12 2019 at 12:32, on Zulip):

yeah pnkfelix good point, I do try to leave comments whenever replacing labels but...

(i know, its easy to forget such things)

pnkfelix (Sep 12 2019 at 12:33, on Zulip):

okay so I had thought #61188 was fixed, but @Matthew Jasper reopened it while I was on leave. So, yeah, hopefully I'll have a chance to look into it soon.

pnkfelix (Sep 12 2019 at 12:34, on Zulip):

P-high 21/41: "ICE when running kcov with proptest as dev-dependency" #60372

pnkfelix (Sep 12 2019 at 12:35, on Zulip):

when I last looked at #60372, it was with an unconfirmed hypothesis that it was a duplicate of #58375.

pnkfelix (Sep 12 2019 at 12:35, on Zulip):

the latter has since been fixed. so I left a comment for the filer (of #60372) asking them if they can check if this bug has similarly been fixed.

pnkfelix (Sep 12 2019 at 12:35, on Zulip):

P-high 22/41: "Compiler panic at Box<Any>" #60363

pnkfelix (Sep 12 2019 at 12:37, on Zulip):

similar to previous issue: there is a possibility that the problem has been fixed, or at least papered-over. Left a comment for the filer (of #60363) asking them to double-check if the problem is still occurring.

pnkfelix (Sep 12 2019 at 12:37, on Zulip):

P-high 23/41: "Stable rustc always panics on arm/musl" #60297

pnkfelix (Sep 12 2019 at 12:38, on Zulip):

I'll go ahead and try to promote this issue in the main channel.

pnkfelix (Sep 12 2019 at 12:39, on Zulip):

P-high 24/41: "ICE while bootstrapping (./x.py build) #2" #60228

pnkfelix (Sep 12 2019 at 12:40, on Zulip):

I'm going to downgrade this to P-medium, as I threatened back in July

pnkfelix (Sep 12 2019 at 12:42, on Zulip):

P-high 25/41: "ICE on higher-trait bounds" #60218

pnkfelix (Sep 12 2019 at 12:43, on Zulip):

yet another bug injected by the old Universes PR (#55517)

pnkfelix (Sep 12 2019 at 12:43, on Zulip):

its assigned to @nikomatsakis . AFAICT that's the best place for it to remain for the time being.

pnkfelix (Sep 12 2019 at 12:44, on Zulip):

P-high 26/41: " Rust 1.34 generates significantly less debug information for libstd functions vs. Rust 1.33" #60020

pnkfelix (Sep 12 2019 at 12:45, on Zulip):

@mw was this resolved by PR #61007 ?

pnkfelix (Sep 12 2019 at 12:47, on Zulip):

P-high 27/41: "Incremental compilation results in linker error when method use is removed" #59535

nikomatsakis (Sep 12 2019 at 12:48, on Zulip):

its assigned to nikomatsakis . AFAICT that's the best place for it to remain for the time being.

thanks, I'm going to try to get on top of these in next week or two

pnkfelix (Sep 12 2019 at 12:48, on Zulip):

Regarding #59535, since I am starting to look into incr-comp issues, I'll unassign @Zoxc here and assign myself. (But @Aaron Turon , if you want this, feel free to work-steal it from me.)

pnkfelix (Sep 12 2019 at 12:49, on Zulip):

P-high 28/41: "Compiler panic with generic-typed nested closures" #59494

mw (Sep 12 2019 at 12:51, on Zulip):

mw was this resolved by PR #61007 ?

Yes, I think that's resolved by that PR

pnkfelix (Sep 12 2019 at 12:52, on Zulip):

regarding #59494, it would be good to get this assigned to someone

pnkfelix (Sep 12 2019 at 12:52, on Zulip):

its been bisected to PR #55986

pnkfelix (Sep 12 2019 at 12:53, on Zulip):

I'm now wishing I had taken more notes from when I was the one investigating it...

pnkfelix (Sep 12 2019 at 12:59, on Zulip):

anyway its an assertion from within the trait select code

pnkfelix (Sep 12 2019 at 12:59, on Zulip):

I'm going to leave the issues label state alone for now; like many other bugs we've looked at today, I do not feel comfortable self-assigning that at this time.

pnkfelix (Sep 12 2019 at 13:00, on Zulip):

P-high 29/41: "NLL compile-time performance regression" #58178

pnkfelix (Sep 12 2019 at 13:00, on Zulip):

I would assume that whatever performance this regressed, we're living with it everywhere now

pnkfelix (Sep 12 2019 at 13:01, on Zulip):

so I'm going to downgrade this from P-high to P-medium.

simulacrum (Sep 12 2019 at 13:01, on Zulip):

I would close

simulacrum (Sep 12 2019 at 13:01, on Zulip):

This isn't something that's worth tracking imo

pnkfelix (Sep 12 2019 at 13:02, on Zulip):

well, it depends on how bad the regression is that this point

simulacrum (Sep 12 2019 at 13:03, on Zulip):

Very slight. See nll page on perf (on mobile so linking is hard)

pnkfelix (Sep 12 2019 at 13:03, on Zulip):

okay, yes, @Matthew Jasper 's work from PR #58347 got the performance delta down to someting more like a 50% hit

pnkfelix (Sep 12 2019 at 13:03, on Zulip):

(we .... don't have an NLL perf page anymore...)

simulacrum (Sep 12 2019 at 13:04, on Zulip):

Hm, I do not recall removing that but perhaps I did

pnkfelix (Sep 12 2019 at 13:04, on Zulip):

well NLL is the default everywhere now

mw (Sep 12 2019 at 13:05, on Zulip):

#60020 is closed :boom:

pnkfelix (Sep 12 2019 at 13:05, on Zulip):

so it may not make sense to have a separate page for it, no?

simulacrum (Sep 12 2019 at 13:05, on Zulip):

Ah, yeah, that would explain it.

pnkfelix (Sep 12 2019 at 13:06, on Zulip):

Anyway I'll start by downgrading to P-medium, and reassigning from @csmoe to myself.

pnkfelix (Sep 12 2019 at 13:07, on Zulip):

I want to double-check the situation; @nnethercote had made some points on https://github.com/rust-lang-nursery/rustc-perf/pull/343 about things still being far too slow, and I want to double check whether that was about rustc performance, or about the benchmark suite itself.

simulacrum (Sep 12 2019 at 13:07, on Zulip):

Sounds good

simulacrum (Sep 12 2019 at 13:08, on Zulip):

(my main point was that I don't think it's worth tracking a performance regression we've already accepted to stable)

pnkfelix (Sep 12 2019 at 13:08, on Zulip):

@simulacrum that, or we need a different way to keep track of such things

simulacrum (Sep 12 2019 at 13:09, on Zulip):

Sure, yes. I've long wanted a "long term areas of interest" or so.

lqd (Sep 12 2019 at 13:09, on Zulip):

(the NLL dashboard was removed in april)

pnkfelix (Sep 12 2019 at 13:09, on Zulip):

the compiler is far too slow, and having some ideas for ways to try to claw back previously lost performance ... okay, yes.

pnkfelix (Sep 12 2019 at 13:10, on Zulip):

P-high 30/41: "Better error message when attempting to instantiate tuple structs with private fields" #58017

pnkfelix (Sep 12 2019 at 13:10, on Zulip):

actually I have to go afk so that I can get home in time to run the actual meeting, so I'm going to pause the P-high traversal here.

nikomatsakis (Sep 12 2019 at 13:11, on Zulip):

Sure, yes. I've long wanted a "long term areas of interest" or so.

yeah

Alexander Regueiro (Sep 12 2019 at 18:21, on Zulip):

@pnkfelix I don't have the time or expertise to tackle that issue, sorry. I was just interested in a solution because it's important to me. I thought you and/or eddyb were tackling it because of the assignments, in any case... Are you okay with it? :-)

nnethercote (Sep 12 2019 at 22:09, on Zulip):

I want to double-check the situation; nnethercote had made some points on https://github.com/rust-lang-nursery/rustc-perf/pull/343 about things still being far too slow, and I want to double check whether that was about rustc performance, or about the benchmark suite itself.

It was about the benchmark itself being too large and thus taking a lot longer than most other benchmarks. The version that landed was cut down significantly so it runs in a reasonable time. TL;DR: problem was solved.

simulacrum (Sep 12 2019 at 22:10, on Zulip):

@pnkfelix ^ so you see it

pnkfelix (Sep 13 2019 at 08:45, on Zulip):

:thumbs_up: ; thanks @simulacrum and @nnethercote

Last update: Nov 21 2019 at 13:05UTC