Stream: t-compiler/wg-prioritization

Topic: I-prioritize #66617 Single file takes 30 minutes to compile


triagebot (May 21 2020 at 03:14, on Zulip):

@WG-prioritization issue #66617 has been requested for prioritization.

Santiago Pastorino (May 27 2020 at 14:17, on Zulip):

this is the last unprioritized I-prioritize

LeSeulArtichaut (May 27 2020 at 14:18, on Zulip):

We'll be able to prioritize more unprioritized regressions then :slight_smile:

LeSeulArtichaut (May 27 2020 at 14:18, on Zulip):

:D

Santiago Pastorino (May 27 2020 at 14:19, on Zulip):

yep

Santiago Pastorino (May 27 2020 at 14:19, on Zulip):

so this one ...

Santiago Pastorino (May 27 2020 at 14:19, on Zulip):

I guess we could try bisecting? or ... hmm the version that provides good result is very old maybe for this?

Santiago Pastorino (May 27 2020 at 14:20, on Zulip):

I'd say that this is P-high but ... hmm

LeSeulArtichaut (May 27 2020 at 14:20, on Zulip):

It's a fairly old regression according to this comment

LeSeulArtichaut (May 27 2020 at 14:20, on Zulip):

1.25

Santiago Pastorino (May 27 2020 at 14:21, on Zulip):

yep I don't remember up to where we can go with nightlies

Santiago Pastorino (May 27 2020 at 14:21, on Zulip):

it's like 180 days or something?

Santiago Pastorino (May 27 2020 at 14:22, on Zulip):

one thing I'd more or less like to have in mind is how many stable versions can we go back by nightlies

Santiago Pastorino (May 27 2020 at 14:23, on Zulip):

it it's 180 days with a 6 week release cycle that would be more or less 4 stable versions

Santiago Pastorino (May 27 2020 at 14:24, on Zulip):

and if all that's true we could test until 1.39 right now

Santiago Pastorino (May 27 2020 at 14:24, on Zulip):

is all this correct?

Santiago Pastorino (May 27 2020 at 14:24, on Zulip):

I don't remember how many days back can we go

LeSeulArtichaut (May 27 2020 at 14:24, on Zulip):

I don't know either :confused:

LeSeulArtichaut (May 27 2020 at 14:25, on Zulip):

But I would like to know, as you said it's an important detail :D

Santiago Pastorino (May 27 2020 at 14:25, on Zulip):

@Pietro Albini how many nightly artifacts do we store?

LeSeulArtichaut (May 27 2020 at 14:25, on Zulip):

You mean bors artifacts?

Pietro Albini (May 27 2020 at 14:25, on Zulip):

exactly 168 days

Pietro Albini (May 27 2020 at 14:25, on Zulip):

so 4 release cycles

Santiago Pastorino (May 27 2020 at 14:25, on Zulip):

:+1:

Santiago Pastorino (May 27 2020 at 14:25, on Zulip):

so we could go back exactly to 1.39

LeSeulArtichaut (May 27 2020 at 14:25, on Zulip):

Great :+1:

Santiago Pastorino (May 27 2020 at 14:26, on Zulip):

not enough to bisect unless we want to do it manually :joy:

Santiago Pastorino (May 27 2020 at 14:26, on Zulip):

anyway P-high? P-medium?

LeSeulArtichaut (May 27 2020 at 14:26, on Zulip):

I'm personally fine with either one

Santiago Pastorino (May 27 2020 at 14:27, on Zulip):

what others think about this one?

Santiago Pastorino (May 27 2020 at 14:27, on Zulip):

I'm a bit more towards P-high but this is also old

Santiago Pastorino (May 27 2020 at 14:28, on Zulip):

hold on, @LeSeulArtichaut have you tried the code?

Yuki Okushi (May 27 2020 at 14:29, on Zulip):

well I don't think this has much of an impact on real-world code since it's quite old, so I'd P-medium

LeSeulArtichaut (May 27 2020 at 14:29, on Zulip):

Santiago Pastorino said:

hold on, LeSeulArtichaut have you tried the code?

What do you mean?

Santiago Pastorino (May 27 2020 at 14:30, on Zulip):

nevermind, it takes a long while to compile :)

LeSeulArtichaut (May 27 2020 at 14:30, on Zulip):

Last time I tried to compile code for a I-compiletime issue it hadn't completed in 2 hours

Santiago Pastorino (May 27 2020 at 14:30, on Zulip):

I was confirming this with latest nightly

LeSeulArtichaut (May 27 2020 at 14:32, on Zulip):

@Yuki Okushi Why did you remove the E-needstest on the issue? I don't doubt there is a reason but I just want to know what it it :big_smile:

LeSeulArtichaut (May 27 2020 at 14:33, on Zulip):

Is E-needstest only for resolved issues?

Santiago Pastorino (May 27 2020 at 14:33, on Zulip):

these kinds of issues make me feel a bit nervous :) to be honest but I guess there are a bunch

Santiago Pastorino (May 27 2020 at 14:33, on Zulip):

@pnkfelix want to know what's your feeling about these issues

Santiago Pastorino (May 27 2020 at 14:33, on Zulip):

would you go with P-high or P-medium?

pnkfelix (May 27 2020 at 14:34, on Zulip):

P-medium I think

Santiago Pastorino (May 27 2020 at 14:34, on Zulip):

in general something like this would qualify as P-high but if there are a ton, about code that's not common in production I guess yeah we should probably go with P-medium

Yuki Okushi (May 27 2020 at 14:34, on Zulip):

LeSeulArtichaut said:

Is E-needstest only for resolved issues?

yeah, I mean so

pnkfelix (May 27 2020 at 14:34, on Zulip):

or at least, it depends

pnkfelix (May 27 2020 at 14:34, on Zulip):

if this were an example of production code, then I might say P-high

pnkfelix (May 27 2020 at 14:35, on Zulip):

but it looks like a microbenchmark designed to illustrate/exacerbate some problem that was probably found in production code

triagebot (May 27 2020 at 14:35, on Zulip):

Issue #66617's prioritization request has been removed.

pnkfelix (May 27 2020 at 14:36, on Zulip):

it also seems like there are a couple distinct follow on tasks

Santiago Pastorino (May 27 2020 at 14:36, on Zulip):

thanks @pnkfelix yeah I agree just wanted to read your reasoning about these in general

Santiago Pastorino (May 27 2020 at 14:36, on Zulip):

what you're saying makes sense :+1:

pnkfelix (May 27 2020 at 14:36, on Zulip):

the E-needstest is relevant here, even if the issue isn't resolved

pnkfelix (May 27 2020 at 14:36, on Zulip):

because the test in question is a benchmark on rustc-perf

pnkfelix (May 27 2020 at 14:36, on Zulip):

which is the kind of thing we can add, in theory, even if the issue isn't resolved

pnkfelix (May 27 2020 at 14:37, on Zulip):

though we may need to trim it down to something that doesn't take 30min to run

Santiago Pastorino (May 27 2020 at 14:37, on Zulip):

agree on that, yeah

pnkfelix (May 27 2020 at 14:37, on Zulip):

(i.e. scale the test down)

pnkfelix (May 27 2020 at 14:37, on Zulip):

also, it seems like there is some debate as to whether the issue itself is resolved or not

Yuki Okushi (May 27 2020 at 14:37, on Zulip):

hm fair enough, re-added

Santiago Pastorino (May 27 2020 at 14:53, on Zulip):

pnkfelix said:

also, it seems like there is some debate as to whether the issue itself is resolved or not

hmm, unsure what comment you've read on the issue make you think this

Santiago Pastorino (May 27 2020 at 14:54, on Zulip):

but I've tried to run it on my machine with latest master and it takes a LOT of time, I needed to abort the execution

Santiago Pastorino (May 27 2020 at 14:54, on Zulip):

and also the latest comment says The newest master seems to have the same issue. I doubt it was fixed.

Santiago Pastorino (May 27 2020 at 14:55, on Zulip):

ahh you're talking about this comment https://github.com/rust-lang/rust/issues/66617#issuecomment-589994433 but yeah, with today's master this is still an issue

Santiago Pastorino (May 27 2020 at 14:56, on Zulip):

commented to clarify a bit

Last update: Jun 05 2020 at 23:15UTC