Stream: t-compiler

Topic: pre-meeting triage 2019-11-21 #54818


pnkfelix (Nov 21 2019 at 13:07, on Zulip):

I will be doing pre-triage in this channel.

pnkfelix (Nov 21 2019 at 13:15, on Zulip):

first up: unprioritized nominated T-compiler issues

pnkfelix (Nov 21 2019 at 13:16, on Zulip):

unpri nom 1/9: "Box is marked as "dereferencable" for the duration of the call" #66600

pnkfelix (Nov 21 2019 at 13:16, on Zulip):

resolving what to do here does seem P-high to me.

pnkfelix (Nov 21 2019 at 13:16, on Zulip):

triage: P-high. Leaving nominated tag on because I think this is worth at least touching on in today meeting (hopefully)

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

unpri nom 2/9: "ICE when rustc failing to write to /dev/null on OSX" #66530

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

this sounds familiar to me

pnkfelix (Nov 21 2019 at 13:18, on Zulip):

oh no, I was thinking of something else; this bug is an artifact of how we derive intermediate build products from the requested output file name

pnkfelix (Nov 21 2019 at 13:20, on Zulip):

does not seem like a terribly high priority thing for us to deal with though.

pnkfelix (Nov 21 2019 at 13:20, on Zulip):

triage: P-medium, assigning self (because the bug author managed to pique my interest by mentioning "rust code minimization"), removing nomination

pnkfelix (Nov 21 2019 at 13:21, on Zulip):

unpri nom 3/9: "matches macro conflicts with matches crate" #66518

centril (Nov 21 2019 at 13:22, on Zulip):

Not sure why that one is t-compiler

pnkfelix (Nov 21 2019 at 13:22, on Zulip):

eek, some scary potential changes to macro hygiene hypothesized here

pnkfelix (Nov 21 2019 at 13:22, on Zulip):

@centril did you read the comments?

pnkfelix (Nov 21 2019 at 13:23, on Zulip):

I wonder how hard this will be to bisect

pnkfelix (Nov 21 2019 at 13:24, on Zulip):

anyway its currently a beta-regression. Lets tag it P-high at least until we've managed to better identify the scope of the bug change in behavior

pnkfelix (Nov 21 2019 at 13:24, on Zulip):

triage: P-high, leaving nominated for now. Leaving unassigned for now.

centril (Nov 21 2019 at 13:25, on Zulip):

Ok so it's a different issue than the expected priority stuff wrt. prelude and whatnot

pnkfelix (Nov 21 2019 at 13:25, on Zulip):

right, exactly

centril (Nov 21 2019 at 13:25, on Zulip):

Let's see what petrochenkov finds then

centril (Nov 21 2019 at 13:25, on Zulip):

Assign to them?

pnkfelix (Nov 21 2019 at 13:25, on Zulip):

its possible someone else could handle the bisection?

pnkfelix (Nov 21 2019 at 13:26, on Zulip):

I'll leave it nominated for now

centril (Nov 21 2019 at 13:26, on Zulip):

Don't see why not

pnkfelix (Nov 21 2019 at 13:26, on Zulip):

if we don't get a volunteer before or at the meeting, then I'll assign to petrochenkov

pnkfelix (Nov 21 2019 at 13:26, on Zulip):

unpri nom 4/9: "ICE matching on a non-[u8] const array" #66501

centril (Nov 21 2019 at 13:27, on Zulip):

This seems related to your work on structural match.

centril (Nov 21 2019 at 13:27, on Zulip):

?

pnkfelix (Nov 21 2019 at 13:28, on Zulip):

that assert sounds frankly wrong?

centril (Nov 21 2019 at 13:28, on Zulip):

I had a look at it yesterday but I don't understand match checking enough

pnkfelix (Nov 21 2019 at 13:29, on Zulip):

Anyway, it seems to me like its only tangentially related to the #[structural_match] work. a slice &[(); n] should be #[structural_match] out of the box.

pnkfelix (Nov 21 2019 at 13:29, on Zulip):

I'll take a look at it

centril (Nov 21 2019 at 13:29, on Zulip):

Cool

pnkfelix (Nov 21 2019 at 13:29, on Zulip):

triage: P-high, removing nomination label, self-assigning.

pnkfelix (Nov 21 2019 at 13:30, on Zulip):

unpri nom 5/9: "ICE 'no entry found for key' in Servo’s style crate" #66487

pnkfelix (Nov 21 2019 at 13:31, on Zulip):

triage: P-high. Removing nomination (petrochenkov already self-assigned.)

pnkfelix (Nov 21 2019 at 13:36, on Zulip):

unpri nom 6/9: "clippy-driver no longer builds after rust-lang/rust#66485" #66486

pnkfelix (Nov 21 2019 at 13:36, on Zulip):

triage: P-medium, removing nomination label.

pnkfelix (Nov 21 2019 at 13:37, on Zulip):

unpri nom 7/9: "ICE: Can't combine functions in the same CGU that reference both a static and an extern static with the same name" #66464

pnkfelix (Nov 21 2019 at 13:37, on Zulip):

triage: has PR; P-high, removing nomination

pnkfelix (Nov 21 2019 at 13:39, on Zulip):

unpri nom 8/9: "Regression in helpful compiler error message when working with lifetimes" #66406

centril (Nov 21 2019 at 13:40, on Zulip):

sounds like a QoL issue; not P-high imo

centril (Nov 21 2019 at 13:40, on Zulip):

but I'm sure estebank will fix it in due time

pnkfelix (Nov 21 2019 at 13:40, on Zulip):

yeah agreed

pnkfelix (Nov 21 2019 at 13:40, on Zulip):

triage: P-medium, removing nomination

pnkfelix (Nov 21 2019 at 13:42, on Zulip):

actually, I agree with @Esteban Küber that this is arguably a dupe of #41343

pnkfelix (Nov 21 2019 at 13:42, on Zulip):

except that this is tagged as a regression due to other artifacts

pnkfelix (Nov 21 2019 at 13:42, on Zulip):

well, I'll let it be for now

pnkfelix (Nov 21 2019 at 13:43, on Zulip):

unpri nom 9/9: "under latest MinGW, cannot link with C code using stdout" #47048

pnkfelix (Nov 21 2019 at 13:44, on Zulip):

I have no idea what's going on here. There's a pretty big comment thread.

pnkfelix (Nov 21 2019 at 13:44, on Zulip):

Leaving unprioritized and nominated.

pnkfelix (Nov 21 2019 at 13:47, on Zulip):

next up: unprioritized beta-regressions

pnkfelix (Nov 21 2019 at 13:47, on Zulip):

unpri β regr 1/3: "NLL Regressions in 1.40" #66517

pnkfelix (Nov 21 2019 at 13:49, on Zulip):

I guess this is P-high

pnkfelix (Nov 21 2019 at 13:49, on Zulip):

I don't know who's going to do the work being requested

pnkfelix (Nov 21 2019 at 13:49, on Zulip):

triage: P-high

centril (Nov 21 2019 at 13:50, on Zulip):

seems like @simulacrum is interested

pnkfelix (Nov 21 2019 at 13:50, on Zulip):

unpri β regr 2/3: "ICE when trying to move from reference in union" #66500

centril (Nov 21 2019 at 13:50, on Zulip):

(personally I suspect these are abandonware and not worth our time, which we could use for other things)

pnkfelix (Nov 21 2019 at 13:50, on Zulip):

seems like simulacrum is interested

maybe, but they did not assign themselves ...

pnkfelix (Nov 21 2019 at 13:51, on Zulip):

regarding #66500, we don't tag things as regressions if they rely on feature-gates, right?

pnkfelix (Nov 21 2019 at 13:51, on Zulip):

oh, I see: the ICE doesn't need the feature gate

pnkfelix (Nov 21 2019 at 13:52, on Zulip):

(I was looking at the next comment that pointed out a slight variation with feature-gate (and &mut) compiles without ICE'ing; which sounds much much worse.)

pnkfelix (Nov 21 2019 at 13:56, on Zulip):

triage: P-high

pnkfelix (Nov 21 2019 at 13:57, on Zulip):

unpri β regr 3/3: "unused_parens triggerts on macro by example code." #66295

simulacrum (Nov 21 2019 at 13:59, on Zulip):

re:NLL breakage, I personally feel like if we are introducing breakage of this scale the least we can do is fix everything or atl least make an effort, regardless of abandonware

simulacrum (Nov 21 2019 at 13:59, on Zulip):

I myself plan to help out, but do not have time to do all 80

pnkfelix (Nov 21 2019 at 13:59, on Zulip):

hmm. I would normally consider #66295 not high priority, but without data about how many crates are impacted (its been flagged a regression), making it P-medium is a little hard to swallow.

centril (Nov 21 2019 at 14:00, on Zulip):

https://github.com/rust-lang/rust/issues/66295#issuecomment-552799156 does warn on stable and older compilers

pnkfelix (Nov 21 2019 at 14:00, on Zulip):

@simulacrum I'll nominate #66517 for discussion at the meeting. Maybe we can collectively agree to divvy up the work in some fair manner

pnkfelix (Nov 21 2019 at 14:00, on Zulip):

@centril that does not jibe with what the bug filer claimed on the bug?

centril (Nov 21 2019 at 14:01, on Zulip):

so it seems that #66295 is an issue wrt. types only? (that's the only new thing that we've changed...)

centril (Nov 21 2019 at 14:01, on Zulip):

@pnkfelix I believe it's like the improper_ctypes thing... it's old behavior manifesting in new places

centril (Nov 21 2019 at 14:02, on Zulip):

"is it a bug" seems like a question which should be answered for all cases (pats, exprs, types) consistently, not just types

centril (Nov 21 2019 at 14:02, on Zulip):

but it seems like a design question?

pnkfelix (Nov 21 2019 at 14:03, on Zulip):

(Personally I'd be inclined to make the lint not fire when its highlighting something inside a macro definition)

centril (Nov 21 2019 at 14:03, on Zulip):

it seems reasonable not to complain about parens in macros

pnkfelix (Nov 21 2019 at 14:03, on Zulip):

lets maybe nominate for a quick straw poll of that idea at the meeting then?

centril (Nov 21 2019 at 14:03, on Zulip):

@pnkfelix seems like we are thinking the same then; but since it's a design question & a lint, it should be brought up in T-Lang (as well...)?

pnkfelix (Nov 21 2019 at 14:04, on Zulip):

okay that's reasonable to me, I'll tag T-lang too

centril (Nov 21 2019 at 14:04, on Zulip):

cheers

pnkfelix (Nov 21 2019 at 14:04, on Zulip):

oh, prioritization

pnkfelix (Nov 21 2019 at 14:04, on Zulip):

lets call it P-high to answer the question

centril (Nov 21 2019 at 14:05, on Zulip):

seems reasonable

pnkfelix (Nov 21 2019 at 14:06, on Zulip):

there are no nightly (ν?) regressions without a P-label

pnkfelix (Nov 21 2019 at 14:07, on Zulip):

next, I'll skim over the nominated T-compiler issues to try to identify cases of labels that should have been removed last week

pnkfelix (Nov 21 2019 at 14:09, on Zulip):

(we discussed "improper_ctypes fires for &mut T, &T, *const T and *mut T (when T: Sized)" #66220 in last week's meeting; I am on hook from lang-team duties to address it with T-lang feedback, but I don't think we'll discuss that one today

pnkfelix (Nov 21 2019 at 14:11, on Zulip):

In any case I'll leave it nominated at least until I post aformentioned T-lang feedback. (Or @centril, would prefer I leave nomination label on even after doing so?)

centril (Nov 21 2019 at 14:12, on Zulip):

not sure; it might be good to discuss it briefly as "last week's stuff"

centril (Nov 21 2019 at 14:12, on Zulip):

mostly to report back to the team

pnkfelix (Nov 21 2019 at 14:14, on Zulip):

removed I-nominated from "Enable incremental rustfmt adoption" #65939

pnkfelix (Nov 21 2019 at 14:14, on Zulip):

we didn't get around to discussing "Some features can no longer be controlled by conditional compilation" #65860 last week despite my best hopes, so I'm leaving that one nominated for today's meeting.

centril (Nov 21 2019 at 14:15, on Zulip):

Yikes; then I need to switch to that one and leave feedback now... but I'm in the middle of providing feedback for something else

pnkfelix (Nov 21 2019 at 14:18, on Zulip):

the thread on that #65860 has become somewhat hard to follow, IMO

pnkfelix (Nov 21 2019 at 14:19, on Zulip):

anyway, that's all the nominations I feel I need to skim. Unfortunately I only managed to cull one from the current todo list

pnkfelix (Nov 21 2019 at 14:19, on Zulip):

I'll try to push these along to get us to the nominations faster today

pnkfelix (Nov 21 2019 at 14:19, on Zulip):

we have 34 open P-high issues

pnkfelix (Nov 21 2019 at 14:20, on Zulip):

nineteen of which are unassigned

pnkfelix (Nov 21 2019 at 14:21, on Zulip):

wow there are 8 beta-nominated PR's

pnkfelix (Nov 21 2019 at 14:22, on Zulip):

five of which are also stable-nominated

pnkfelix (Nov 21 2019 at 14:22, on Zulip):

might be hard to get through 8 beta-nominations and 11 I-nominated issues

pnkfelix (Nov 21 2019 at 14:22, on Zulip):

we will see

centril (Nov 21 2019 at 14:23, on Zulip):

@pnkfelix did you forget to leave beta approvals from last week?

pnkfelix (Nov 21 2019 at 14:23, on Zulip):

oh good point!

pnkfelix (Nov 21 2019 at 14:23, on Zulip):

let me go check

pnkfelix (Nov 21 2019 at 14:25, on Zulip):

yes yes yes thank goodness

centril (Nov 21 2019 at 14:25, on Zulip):

:slight_smile:

Esteban Küber (Nov 21 2019 at 14:26, on Zulip):

Reminder that I'd like to touch on #53081, if we have the time

Esteban Küber (Nov 21 2019 at 14:26, on Zulip):

But less critical than the regressions

pnkfelix (Nov 21 2019 at 14:30, on Zulip):

@Esteban Küber yep, its one of the 11 nominated issues. I'll try to put it at the front of the queue

lqd (Nov 21 2019 at 14:35, on Zulip):

@simulacrum what do you want to do for the abandoned ones (or GH repos which are just tests) ? I'd like to help but the ones I'm looking at rn have either been deprecated, archived on GH, etc

simulacrum (Nov 21 2019 at 14:36, on Zulip):

If they're explicitly abandoned then that's fine

simulacrum (Nov 21 2019 at 14:36, on Zulip):

Basically - we should put in the time

simulacrum (Nov 21 2019 at 14:36, on Zulip):

And not just say "too bad"

Esteban Küber (Nov 21 2019 at 14:39, on Zulip):

There's code out there that is not visible to us, if there are regressions in the open it means there's a non insignificant chance that closed repos will also hit it.

Esteban Küber (Nov 21 2019 at 14:39, on Zulip):

Not even a clean crater run means a safe change, it only gives us _some_ confidence.

lqd (Nov 21 2019 at 14:43, on Zulip):

@simulacrum sure, I'll add what I saw to #66517

centril (Nov 21 2019 at 14:45, on Zulip):

Basically - we should put in the time

(I personally don't agree with this; I think there's also a question of fairness towards all the unpaid FOSS contributors in rust-lang/rust -- I don't think we/they should bear all the cost, especially when crates are largely unused, and when the migration period has been so long)

centril (Nov 21 2019 at 14:47, on Zulip):

(of course, if it's important to you then by all means... ^^)

Last update: Dec 12 2019 at 00:55UTC