Stream: t-compiler

Topic: pre-meeting triage 2019-12-19 #54818


pnkfelix (Dec 19 2019 at 13:16, on Zulip):

I will be doing pre-triage in this channel.

pnkfelix (Dec 19 2019 at 13:34, on Zulip):

first up: unprioritized nominations

pnkfelix (Dec 19 2019 at 13:34, on Zulip):

unpri nom 1/7: "missing libstdc++-6.dll in beta-x86_64-pc-windows-gnu (1.41.0-beta.1)" #67408

pnkfelix (Dec 19 2019 at 13:35, on Zulip):

has PR

pnkfelix (Dec 19 2019 at 13:35, on Zulip):

#67408: P-high, removing nomination

pnkfelix (Dec 19 2019 at 13:42, on Zulip):

also, just so people are aware: PR #67410 (the fix for the aforementioned issue) has already been beta-nominated and insta-beta-accepted (presumably due to release pressure). Thanks @simulacrum !

simulacrum (Dec 19 2019 at 13:45, on Zulip):

Uh, not really - just not really any need to consult anyone (Alex and I are fine with backport I believe and only maintainers of bootstrap today)

pnkfelix (Dec 19 2019 at 13:47, on Zulip):

oh its not release pressure? Isn't there a imminent beta promotion ?

pnkfelix (Dec 19 2019 at 13:56, on Zulip):

unpri nom 2/7: "const_if_match ICE" #67405

pnkfelix (Dec 19 2019 at 13:58, on Zulip):

triage: stable-to-beta regression. P-high. Leaving nominated to try to ensure we give it attention at todays meeting.

pnkfelix (Dec 19 2019 at 14:02, on Zulip):

unpri nom 3/7: "ICE: unexpected ty: [type error]" #67377

pnkfelix (Dec 19 2019 at 14:06, on Zulip):

I wrote a comment. I think this is only P-medium, because the common issue with the examples is that a syntax error in the source (which is reported in a diagnostic) is somehow causing an ICE downstream after the parser does recovery

pnkfelix (Dec 19 2019 at 14:07, on Zulip):

unpri nom 4/7: "ICE on unused generic in struct" #67375

pnkfelix (Dec 19 2019 at 14:09, on Zulip):

wow that one is pretty wild looking

pnkfelix (Dec 19 2019 at 14:10, on Zulip):

I'm willing to call this P-high, just because there is no diagnostic info apart from the ICE, and its message is not very helpful for a non-rustc dev

pnkfelix (Dec 19 2019 at 14:10, on Zulip):

#67375: P-high, removing nomination.

pnkfelix (Dec 19 2019 at 14:11, on Zulip):

unpri nom 5/7: "Performance regression in nightly-2019-12-14" #67331

pnkfelix (Dec 19 2019 at 14:11, on Zulip):

cannot act on this yet since there is no example code yet for us to work with

pnkfelix (Dec 19 2019 at 14:12, on Zulip):

I'm going to remove the nomination and ask them to renominate after providing an example we can use

pnkfelix (Dec 19 2019 at 14:14, on Zulip):

unpri nom 6/7: "Casting or adding type ascription to panic!() triggers unreachable_code" #67227

simulacrum (Dec 19 2019 at 14:18, on Zulip):

No beta promotion has already happened

pnkfelix (Dec 19 2019 at 14:23, on Zulip):

for #67227, I'm not sure whether there is anything we can do here yet. We could investigate trying to make the unreachable_code lint treat coercions/casts specially

pnkfelix (Dec 19 2019 at 14:24, on Zulip):

in any case, #67227 doesn't seem like a high priority item to address.

pnkfelix (Dec 19 2019 at 14:24, on Zulip):

I'm going to triage it as P-medium but leave it nominated.

centril (Dec 19 2019 at 14:26, on Zulip):

(T-Lang still needs to make a decision)

centril (Dec 19 2019 at 14:26, on Zulip):

(attendance was low in prev mtg)

pnkfelix (Dec 19 2019 at 14:26, on Zulip):

gotcha

pnkfelix (Dec 19 2019 at 14:27, on Zulip):

unpri nom: "Replace our fragile safety scheme around erroneous constants" #67191

pnkfelix (Dec 19 2019 at 14:28, on Zulip):

triage: P-high to resolve desired semantics and implementation strategy here. (I don't know what priority to assign to the actual implementation work yet though...)

pnkfelix (Dec 19 2019 at 14:29, on Zulip):

leaving #67191 nominated because I think T-lang may need to revisit it when they/we have a proper quorum

pnkfelix (Dec 19 2019 at 14:31, on Zulip):

that's all the unprioritized nominations

pnkfelix (Dec 19 2019 at 14:31, on Zulip):

there are no beta-regressions without a P-label

pnkfelix (Dec 19 2019 at 14:32, on Zulip):

there are no nightly-regressions without a P-label

pnkfelix (Dec 19 2019 at 14:32, on Zulip):

I'll skim over the nominations, to 1. see if there's cases where I can remove the nomination (e.g. if its alreay been addressed) and 2. to curate the list of nominations I would like to ensure we address today

pnkfelix (Dec 19 2019 at 14:33, on Zulip):

there are 10 open nominations

pnkfelix (Dec 19 2019 at 14:35, on Zulip):

I'm going to put "./x.py check failed if incremental builds enabled" #58633 on the list again for today: @simulacrum and I were musing last night about whether we should just remove the unused_attributes lint entirely. (I believe it has served its original purpose.)

pnkfelix (Dec 19 2019 at 14:38, on Zulip):

@mati865 are you around for today's meeting? I'd like someone to advocate for movement on "under latest MinGW, cannot link with C code using stdout" #47048

pnkfelix (Dec 19 2019 at 14:41, on Zulip):

it sounds like @mati865 is at this point [blocked by issues with the old version of mingw-w64 on Rust's CI. Does that make this a T-infra issue?

pnkfelix (Dec 19 2019 at 14:41, on Zulip):

@simulacrum @Pietro Albini ^ ?

simulacrum (Dec 19 2019 at 14:42, on Zulip):

Someone needs to make the call it we're to upgrade

simulacrum (Dec 19 2019 at 14:42, on Zulip):

I'd guess that's compiler team

pnkfelix (Dec 19 2019 at 14:42, on Zulip):

okay I'll put this on today's agenda too then

centril (Dec 19 2019 at 14:45, on Zulip):

whether we should just remove the unused_attributes lint entirely. (I believe it has served its original purpose.)

I'd like to see a list of cases where the lint would could be triggered to evaluate whether the lang team should decide to remove it or not.
I know it would trigger on e.g. #[derive()].

pnkfelix (Dec 19 2019 at 14:46, on Zulip):

@centril maybe I chose too strong a phrasing

pnkfelix (Dec 19 2019 at 14:46, on Zulip):

it isn't necessarily that the lint is totally useless

centril (Dec 19 2019 at 14:47, on Zulip):

@pnkfelix I don't really have an opinion whether it is or isn't at this point :slight_smile:

centril (Dec 19 2019 at 14:47, on Zulip):

(shocking, I know... Centril doesn't have an opinion? ^^)

pnkfelix (Dec 19 2019 at 14:47, on Zulip):

the problem is that its current architecture, based on updating some global state, has not been updated to work with incremental compilation

centril (Dec 19 2019 at 14:47, on Zulip):

right yeah; mark_used (https://doc.rust-lang.org/nightly/nightly-rustc/syntax/attr/fn.mark_used.html) & friends are pretty ugly from an infra POV

pnkfelix (Dec 19 2019 at 14:48, on Zulip):

and so we are in this situation where we need to decide whether the lint is providing enough of a benefit that we should invest the effort to make it actually work properly with incremental comp. My brief review o the code last night led me to think that it was not a trivial fix

centril (Dec 19 2019 at 14:49, on Zulip):

Right, I can see how that'd be the case that it plays badly with incr.comp -- I'm less sure about the benefit side of things

pnkfelix (Dec 19 2019 at 14:49, on Zulip):

for example, could T-compiler decide to remove it or disable it across the board, without lang team approval, if we also filed a bug saying "this thing was buggy; T-lang can decide whether its worth reenabling" ?

pnkfelix (Dec 19 2019 at 14:50, on Zulip):

(@simulacrum did have at least one idea on alternative implementation strategies that might not be too hard to implement, but the proposal, namely a bit of mutable state in the AST itself, scared the bejeeezus out of me.)

centril (Dec 19 2019 at 14:50, on Zulip):

I'd say: let's discuss it on T-Lang today; and then we don't need to go back to a t-compiler meeting again, someone can "just" do the removal if we decide to do it

pnkfelix (Dec 19 2019 at 14:51, on Zulip):

okay, I'm okay with that. I'll remove it from T-compiler agenda for today then

centril (Dec 19 2019 at 14:51, on Zulip):

namely a bit of mutable state in the AST itself, scared the bejeeezus out of me.

This will give eddyb a heart attack ;)

centril (Dec 19 2019 at 14:51, on Zulip):

(and me also)

pnkfelix (Dec 19 2019 at 14:51, on Zulip):

(or was your implication that it should nonetheless stay on T-compiler agenda for today ...?)

centril (Dec 19 2019 at 14:51, on Zulip):

if you think it's useful

centril (Dec 19 2019 at 14:52, on Zulip):

maybe someone comes up with an alternative alternative strategy on the mtg

pnkfelix (Dec 19 2019 at 14:52, on Zulip):

eh... we might have a lot of stuff to get through

simulacrum (Dec 19 2019 at 14:52, on Zulip):

I would not worry about it for today

simulacrum (Dec 19 2019 at 14:53, on Zulip):

not that high priority :)

centril (Dec 19 2019 at 14:53, on Zulip):

(can we use a side-table via ParseSess?)

simulacrum (Dec 19 2019 at 14:54, on Zulip):

I think yes -- but the side table presumably doesn't solve the problem of this state not being incremental-ready

mati865 (Dec 19 2019 at 14:54, on Zulip):

mati865 are you around for today's meeting? I'd like someone to advocate for movement on "under latest MinGW, cannot link with C code using stdout" #47048

Sorry, just got the email.
I think it requires T-compiler fix first (I have WIP in https://rust-lang.zulipchat.com/#narrow/stream/131828-t-compiler/topic/.5BWIP.5D.20Fixing.20MinGW.20link.20errors.20.2347048) and after it's resolved it'll be purely T-infra issue (updating mingw-w64).

pnkfelix (Dec 19 2019 at 14:54, on Zulip):

FYI I'm removing nomination from "Linking issue with Rust 1.37.0" #64340

mati865 (Dec 19 2019 at 14:54, on Zulip):

I'm preparing newer version to open PR today.

simulacrum (Dec 19 2019 at 14:55, on Zulip):

well, I don't consider that a T-infra issue

simulacrum (Dec 19 2019 at 14:55, on Zulip):

i.e., we're not updating because we don't know whether it's safe

simulacrum (Dec 19 2019 at 14:55, on Zulip):

(i.e., is that a compat break for someone)

pnkfelix (Dec 19 2019 at 14:55, on Zulip):

I thought you just said T-compiler should make call ?

simulacrum (Dec 19 2019 at 14:55, on Zulip):

right, yeah, that's my point

pnkfelix (Dec 19 2019 at 14:55, on Zulip):

oh oh, right, I see.

pnkfelix (Dec 19 2019 at 14:55, on Zulip):

but after T-compiler makes the call

pnkfelix (Dec 19 2019 at 14:56, on Zulip):

then the actual work is T-infra, right?

pnkfelix (Dec 19 2019 at 14:56, on Zulip):

(I'm not saying that's fair. Just that it reflects reality...)

simulacrum (Dec 19 2019 at 14:56, on Zulip):

Sure, yes

simulacrum (Dec 19 2019 at 14:57, on Zulip):

though I envision the work being pretty simple -- the investigation into whether we want to do it is the hard bit

pnkfelix (Dec 19 2019 at 14:58, on Zulip):

okay I'm done with my nomination skim

pnkfelix (Dec 19 2019 at 14:58, on Zulip):

time to put up the beta-noms list

pnkfelix (Dec 19 2019 at 14:59, on Zulip):

two beta-noms; I'll put them on the hackmd I'm drafting.

Last update: May 26 2020 at 10:55UTC