Stream: t-compiler

Topic: #63370 beta backport


Pietro Albini (Aug 12 2019 at 15:30, on Zulip):

Hey @T-compiler! The promotion from beta to stable is going to happen really soon.

Pietro Albini (Aug 12 2019 at 15:30, on Zulip):

do you care about backporting https://github.com/rust-lang/rust/pull/63370 ?

simulacrum (Aug 12 2019 at 15:34, on Zulip):

We're going to go ahead without it in 10 minutes or so, fwiw, since it seems non-critical and very late in the cycle.

nagisa (Aug 12 2019 at 15:43, on Zulip):

Won't backport I think.

nagisa (Aug 12 2019 at 15:43, on Zulip):

Doing so would violate the process for doing so.

Pietro Albini (Aug 12 2019 at 15:44, on Zulip):

I'm also dropping https://github.com/rust-lang/rust/pull/62982 as a backport

Pietro Albini (Aug 12 2019 at 15:44, on Zulip):

it just doesn't apply on beta

nagisa (Aug 12 2019 at 15:44, on Zulip):

Ask @oli to backport. This is a stable-to-beta regression fix

oli (Aug 12 2019 at 15:45, on Zulip):

will do

Pietro Albini (Aug 12 2019 at 15:46, on Zulip):

@simulacrum ok I'll wait for oli's backport PR to send the promotion

simulacrum (Aug 12 2019 at 15:47, on Zulip):

sounds good

simulacrum (Aug 12 2019 at 17:23, on Zulip):

@oli Please ping me and Pietro when backport is ready

oli (Aug 12 2019 at 17:23, on Zulip):

Like this very second

oli (Aug 12 2019 at 17:23, on Zulip):

i literally just switched to the browser

oli (Aug 12 2019 at 17:23, on Zulip):

because the push went through

simulacrum (Aug 12 2019 at 17:24, on Zulip):

heh okay

simulacrum (Aug 12 2019 at 17:25, on Zulip):

we'll be rolling it up immediately so if just a branch I can pull from is fine

simulacrum (Aug 12 2019 at 17:25, on Zulip):

@oli no need to open a PR

oli (Aug 12 2019 at 17:25, on Zulip):

https://github.com/rust-lang/rust/commit/de114a54c0a43dbbcfc5fd53fc202c5249104c72

oli (Aug 12 2019 at 17:26, on Zulip):

I just tried to open a PR but there are weird commits in between even though I just rebased

oli (Aug 12 2019 at 17:26, on Zulip):

anyway, it's just this commit

oli (Aug 12 2019 at 17:26, on Zulip):

I removed all the cleanup part from the PR

oli (Aug 12 2019 at 17:26, on Zulip):

and just left the fix commit

simulacrum (Aug 12 2019 at 17:26, on Zulip):

perfect

simulacrum (Aug 12 2019 at 17:46, on Zulip):

@oli That commit doesn't appear to merge cleanly with beta?

simulacrum (Aug 12 2019 at 17:50, on Zulip):

cc @RalfJ perhaps you might have a chance to help backport the patch?

RalfJ (Aug 12 2019 at 17:59, on Zulip):

what which patch? I don't think I have seen the PR in the topic description before

RalfJ (Aug 12 2019 at 18:00, on Zulip):

oh you mean oli's patch

RalfJ (Aug 12 2019 at 18:00, on Zulip):

but isnt a new stable released in <1 week? and that one needs an urgent backport or what is happening?

simulacrum (Aug 12 2019 at 18:01, on Zulip):

Yes, we need a backport to it

simulacrum (Aug 12 2019 at 18:01, on Zulip):

stable and beta branches are the same right now (I've already force pushed beta to stable)

simulacrum (Aug 12 2019 at 18:01, on Zulip):

https://github.com/rust-lang/rust/pull/63498

RalfJ (Aug 12 2019 at 18:01, on Zulip):

well that PR also introduced a bug

RalfJ (Aug 12 2019 at 18:01, on Zulip):

so we shouldnt just plain backport it I imagine

RalfJ (Aug 12 2019 at 18:02, on Zulip):

but I could try backporting the final result...

simulacrum (Aug 12 2019 at 18:03, on Zulip):

that'd be great

RalfJ (Aug 12 2019 at 18:03, on Zulip):

https://github.com/rust-lang/rust/pull/63498

you already force-pushed but this is about doing that switch? I dont get it but whatever^^

simulacrum (Aug 12 2019 at 18:03, on Zulip):

the release process is force push and then file a PR against stable switching things over and last-minute backports

RalfJ (Aug 12 2019 at 18:04, on Zulip):

okay so I should PR against stable? as beta will be force-pushd over on Tuesday anyway?

RalfJ (Aug 12 2019 at 18:04, on Zulip):

I guess this will destroy my LLVM build cache for one of my checkouts? oh well^^

simulacrum (Aug 12 2019 at 18:05, on Zulip):

It'll just be cherry-picked into my PR anyway, so a commit is fine

RalfJ (Aug 12 2019 at 18:05, on Zulip):

I have to admit I feel a bit uncomfortable with just how little testing this will see

simulacrum (Aug 12 2019 at 18:05, on Zulip):

@nagisa seemed to feel this was important enough to not skip it

simulacrum (Aug 12 2019 at 18:06, on Zulip):

I'm not sure how high impact this is.

simulacrum (Aug 12 2019 at 18:06, on Zulip):

I don't recall Crater hitting it, FWIW.

nagisa (Aug 12 2019 at 18:07, on Zulip):

I don't want to have stable-to-beta regressions to become stable-to-stable if we can avoid it

nagisa (Aug 12 2019 at 18:07, on Zulip):

having a compiler that is broken for a single version complicates things

simulacrum (Aug 12 2019 at 18:07, on Zulip):

We don't have a patch at this point and I get the impression that maybe making one is somewhat hard?

simulacrum (Aug 12 2019 at 18:07, on Zulip):

Especially in that it'll receive ~no testing

simulacrum (Aug 12 2019 at 18:08, on Zulip):

beyond regular rust-lang/rust CI

nagisa (Aug 12 2019 at 18:08, on Zulip):

well if we cannot avoid it, I cannot really force anybody to do anything ;)

simulacrum (Aug 12 2019 at 18:08, on Zulip):

To be clear, I feel comfortable enough presuming the patch is similar to the one in nightly

simulacrum (Aug 12 2019 at 18:09, on Zulip):

@RalfJ seemed to be at least mildly concerned though (and they're more familiar with that area of the compiler than I)

RalfJ (Aug 12 2019 at 18:09, on Zulip):

We don't have a patch at this point and I get the impression that maybe making one is somewhat hard?

no not really hard

RalfJ (Aug 12 2019 at 18:09, on Zulip):

I have a patch, build-testing it now

RalfJ (Aug 12 2019 at 18:09, on Zulip):

but this code keeps surprising me with interactions

RalfJ (Aug 12 2019 at 18:10, on Zulip):

I feel more comfortable making it similar to how the method looks these days than similar to that original patch

RalfJ (Aug 12 2019 at 18:10, on Zulip):

as that's fairly well-structured and also well-tested

simulacrum (Aug 12 2019 at 18:11, on Zulip):

That would be fine with me, though I presume that means a larger patch? I think the size doesn't matter too much (though someone's review, ideally not mine, would be great)

simulacrum (Aug 12 2019 at 18:11, on Zulip):

I don't feel comfortable approving it myself

RalfJ (Aug 12 2019 at 18:13, on Zulip):

@simulacrum doesnt really make it larger, no. here's a preview: https://github.com/RalfJung/rust/commit/62139840b33ab73f1135f47116aa6196a49e6d99

RalfJ (Aug 12 2019 at 18:13, on Zulip):

still build-testing

RalfJ (Aug 12 2019 at 18:15, on Zulip):

I don't understand how oli's patch can still have InboundsCheck in it...

simulacrum (Aug 12 2019 at 18:15, on Zulip):

I think it targets master? Not sure.

RalfJ (Aug 12 2019 at 18:16, on Zulip):

no it targets something way older

RalfJ (Aug 12 2019 at 18:16, on Zulip):

InboundsCheck is gone for a while

RalfJ (Aug 12 2019 at 18:16, on Zulip):

and current stable branch doesnt seem to have it any more

simulacrum (Aug 12 2019 at 18:16, on Zulip):

hm, maybe previous stable then, not sure

simulacrum (Aug 12 2019 at 18:16, on Zulip):

1.36.0 tag

simulacrum (Aug 12 2019 at 18:16, on Zulip):

(I recently force pushed)

RalfJ (Aug 12 2019 at 18:16, on Zulip):

I guess I should copy the testcase as well

RalfJ (Aug 12 2019 at 18:28, on Zulip):

@simulacrum https://github.com/RalfJung/rust/commit/331e09b143aebfcf82dc1f9b69b31ee0083cbf0b at least builds. it'll be a bit until I know more, LLVM is building.

simulacrum (Aug 12 2019 at 18:31, on Zulip):

@RalfJ do you know who a good reviewer would be at this hour?

RalfJ (Aug 12 2019 at 18:35, on Zulip):

@oli , maybe @eddyb

RalfJ (Aug 12 2019 at 18:35, on Zulip):

but I guess oli isnt available or they would have done the backport?

simulacrum (Aug 12 2019 at 18:37, on Zulip):

that's my impression yes

RalfJ (Aug 12 2019 at 18:39, on Zulip):

the bus factor of that code is not very high :/

RalfJ (Aug 12 2019 at 18:58, on Zulip):

@simulacrum tests are looking good

simulacrum (Aug 12 2019 at 18:59, on Zulip):

okay I've pinged folks on the PR but I'm going to (attempt to) land it in the meanwhile, presuming the code is good, and we can re-make artifacts if we need to

simulacrum (Aug 12 2019 at 19:00, on Zulip):

we have 2 hours before the currently testing PR bounces or lands too

RalfJ (Aug 12 2019 at 19:14, on Zulip):

got the latest commit in https://github.com/RalfJung/rust/tree/stable/miri-static-size-and-align but that's what I posted above

RalfJ (Aug 12 2019 at 20:55, on Zulip):

@simulacrum I ran the Miri test suite from that time on the branch with my patch and it passes. That's usually a good stress test for the Miri engine (better than just the CTFE tests as many features are not allowed in CTFE, so some code paths are hard to trigger).

RalfJ (Aug 12 2019 at 20:56, on Zulip):

I can't think of anything else to do, and I can feel my concentration slip, so I'll call this done for now ;)

simulacrum (Aug 12 2019 at 20:56, on Zulip):

more good news, thanks!

simulacrum (Aug 12 2019 at 20:56, on Zulip):

I'd still like someone to give it a review given immediate promotion to stable but given all that it seems likely we should be good

centril (Aug 12 2019 at 20:58, on Zulip):

cc @eddyb ^---

RalfJ (Aug 12 2019 at 20:58, on Zulip):

yeah getting an expert review would be good. maybe you can make @eddyb do it. ;)

RalfJ (Aug 12 2019 at 20:58, on Zulip):

I'll head off for the night. good night all!

eddyb (Aug 13 2019 at 07:16, on Zulip):

oops, sorry I wasn't around

simulacrum (Aug 13 2019 at 10:51, on Zulip):

@eddyb The PR has been merged, but mind taking a look at https://github.com/RalfJung/rust/commit/331e09b143aebfcf82dc1f9b69b31ee0083cbf0b?

eddyb (Aug 13 2019 at 10:59, on Zulip):

@simulacrum LGTM but maybe also ask @oli

simulacrum (Aug 13 2019 at 10:59, on Zulip):

okay, thanks!

oli (Aug 13 2019 at 11:00, on Zulip):

yes

oli (Aug 13 2019 at 11:00, on Zulip):

sorry, I checked it but forgot to tell anyone

simulacrum (Aug 13 2019 at 11:07, on Zulip):

okay, perfect

simulacrum (Aug 13 2019 at 11:07, on Zulip):

cc @Pietro Albini looks like we don't need to reprepare stable

Pietro Albini (Aug 13 2019 at 11:25, on Zulip):

:thumbs_up:

Last update: Nov 16 2019 at 01:15UTC