Stream: t-compiler/meetings

Topic: [weekly meeting] 2020-10-01 #54818


Santiago Pastorino (Sep 30 2020 at 22:49, on Zulip):

Hi @T-compiler/meeting; the triage meeting will happen tomorrow at

Santiago Pastorino (Sep 30 2020 at 22:49, on Zulip):

WG-prioritization has done pre-triage in #t-compiler/wg-prioritization/alerts

Santiago Pastorino (Sep 30 2020 at 22:50, on Zulip):

@WG-prioritization has prepared the meeting agenda

Santiago Pastorino (Sep 30 2020 at 22:50, on Zulip):

We will have checkins from @WG-rls2.0 and @WG-self-profile

Santiago Pastorino (Sep 30 2020 at 22:51, on Zulip):

@matklad do you have something you want to share about @WG-rls2.0?

Santiago Pastorino (Sep 30 2020 at 22:51, on Zulip):

@Wesley Wiser do you have something you want to share about @WG-self-profile?

Santiago Pastorino (Oct 01 2020 at 12:19, on Zulip):

Hi @T-compiler/meeting, triage meeting will be starting in 1 hour and 41 minutes

Santiago Pastorino (Oct 01 2020 at 12:19, on Zulip):

Check out the meeting agenda

Santiago Pastorino (Oct 01 2020 at 12:20, on Zulip):

Santiago Pastorino said:

We will have checkins from @WG-rls2.0 and @WG-self-profile

@matklad @Wesley Wiser friendly reminder of :point_up: in case you want to share something :heart:

pnkfelix (Oct 01 2020 at 13:59, on Zulip):

Hi @T-compiler/meeting! Add a :wave: emoji to show you're here :)

pnkfelix (Oct 01 2020 at 14:00, on Zulip):

we will start off with 5 minutes for ...

Announcements

pnkfelix (Oct 01 2020 at 14:00, on Zulip):
pnkfelix (Oct 01 2020 at 14:00, on Zulip):
pnkfelix (Oct 01 2020 at 14:00, on Zulip):
pnkfelix (Oct 01 2020 at 14:00, on Zulip):
pnkfelix (Oct 01 2020 at 14:01, on Zulip):
pnkfelix (Oct 01 2020 at 14:01, on Zulip):
Santiago Pastorino (Oct 01 2020 at 14:01, on Zulip):

I'd love to hear if this :point_up: is useful (given that's not T-compiler)

pnkfelix (Oct 01 2020 at 14:02, on Zulip):

(oop's; I meant those to be two separate messages but I overlooked the bulleting. My gut told me that list looked too long!)

Santiago Pastorino (Oct 01 2020 at 14:02, on Zulip):

Santiago Pastorino said:

I'd love to hear if this :point_up: is useful (given that's not T-compiler)

please let us know in #t-compiler/wg-prioritization > finalized fcps for other teams

Wesley Wiser (Oct 01 2020 at 14:03, on Zulip):

rust#75914 and rust#75991 both only need one more person to check their box before they go into final comment period.

Wesley Wiser (Oct 01 2020 at 14:03, on Zulip):

:smile:

simulacrum (Oct 01 2020 at 14:03, on Zulip):
simulacrum (Oct 01 2020 at 14:04, on Zulip):

(also on comparison pages in a secondary section, unclear how we'll combine the two as of yet)

simulacrum (Oct 01 2020 at 14:05, on Zulip):
pnkfelix (Oct 01 2020 at 14:06, on Zulip):

simulacrum said:

On that chart, the rustc_middle line seems like it settled at ~200 seconds to build. But there is one data point where it took 100 seconds. Do we know why that point was so low?

simulacrum (Oct 01 2020 at 14:07, on Zulip):

I would ignore the beginning, that was experimentation with different -jN levels and such

pnkfelix (Oct 01 2020 at 14:07, on Zulip):

ah gotcha

pnkfelix (Oct 01 2020 at 14:07, on Zulip):

would it be worthwhile to remove those data points from the presentation then?

pnkfelix (Oct 01 2020 at 14:07, on Zulip):

(or will that happen naturally?)

simulacrum (Oct 01 2020 at 14:07, on Zulip):

probably, yeah, just haven't gotten around to editing it, but it'll happen naturally in a month or so

pnkfelix (Oct 01 2020 at 14:08, on Zulip):

okay then maybe not worth stressing about.

pnkfelix (Oct 01 2020 at 14:08, on Zulip):

unless you see other people asking the same Q I did. :)

pnkfelix (Oct 01 2020 at 14:08, on Zulip):

WG checkins

pnkfelix (Oct 01 2020 at 14:09, on Zulip):

@matklad if you have checkin info from @WG-rls2.0 , feel free to post it here or in the hackmd agenda

pnkfelix (Oct 01 2020 at 14:09, on Zulip):

@WG-self-profile checkin by @Wesley Wiser:

We've had quite a lot of activity recently!
- We released measureme 0.8 today :tada:
- Release highlights:
- We have a new storage format contributed by mw that records all of our data in one file instead of three
- @_wesleywiser made some UI tweaks to the summarize tool so it's a bit nicer to use. The main one is that we no longer show columns for data that isn't applicable like parallel query blocking time if the compiler wasn't built with parallel support.
- @_simulacrum made it possible to run the analysis tools from perf.rlo without writing profile data to disk.
> - Several contributors (@_workingjubilee, @_Aaron1011, @_wesleywiser) updated and removed unneeded dependencies :medal:
- @_mw has been working on additional improvements so that we can record query keys when building rustc_middle. Currently, rustc_middle is so large, it exhausts the address space we use for recording string data. rust-lang/measureme#137 will resolve that and allow us to profile even extremely large crates with query keys enabled. This is a breaking change so 0.9.0 (or perhaps 9.0) is likely going to ship soon.
- @_eddyb continues to work on adding support for reading hardware performance counters like perf does instead of using std::time::Instant for much more precise timings.

pnkfelix (Oct 01 2020 at 14:10, on Zulip):

@eddyb (or @Wesley Wiser ) Will the HW perf counters also decrease noise in addition to increasing precision?

pnkfelix (Oct 01 2020 at 14:11, on Zulip):

(perhaps that will depend on which counters are used...)

eddyb (Oct 01 2020 at 14:11, on Zulip):

the additional "precision" specifically means less noise

pnkfelix (Oct 01 2020 at 14:11, on Zulip):

cool, thanks

eddyb (Oct 01 2020 at 14:11, on Zulip):

or rather, 0 noise :sweat_smile:

eddyb (Oct 01 2020 at 14:11, on Zulip):

(at least on some platforms)

pnkfelix (Oct 01 2020 at 14:12, on Zulip):

Beta-nominations

T-compiler

pnkfelix (Oct 01 2020 at 14:12, on Zulip):
pnkfelix (Oct 01 2020 at 14:13, on Zulip):

yeah I should have just unilaterally beta-approved this one

pnkfelix (Oct 01 2020 at 14:13, on Zulip):

beta approved

pnkfelix (Oct 01 2020 at 14:13, on Zulip):

libs-impl

T-rustdoc

pnkfelix (Oct 01 2020 at 14:13, on Zulip):

Stable-nominations

T-compiler

libs-impl

T-rustdoc

pnkfelix (Oct 01 2020 at 14:13, on Zulip):

PRs S-waiting-on-team

T-compiler

libs-impl

pnkfelix (Oct 01 2020 at 14:13, on Zulip):

Issues of Note

Short Summary

pnkfelix (Oct 01 2020 at 14:13, on Zulip):

P-critical

T-compiler

pnkfelix (Oct 01 2020 at 14:14, on Zulip):
pnkfelix (Oct 01 2020 at 14:15, on Zulip):

looks like @Aaron Hill was assigned as a reviewer on the nicer solution too

Santiago Pastorino (Oct 01 2020 at 14:15, on Zulip):

as the last line says a possible better solution was proposed by the person that was going to review @Aaron Hill's PR

pnkfelix (Oct 01 2020 at 14:15, on Zulip):

this seems like its under control, in any case

pnkfelix (Oct 01 2020 at 14:15, on Zulip):
pnkfelix (Oct 01 2020 at 14:16, on Zulip):

(we just approved the backport on this one)

Santiago Pastorino (Oct 01 2020 at 14:16, on Zulip):

with the just approved backport this is solved

simulacrum (Oct 01 2020 at 14:16, on Zulip):

yep, and backport already happened actually so this is good

pnkfelix (Oct 01 2020 at 14:16, on Zulip):
Wesley Wiser (Oct 01 2020 at 14:16, on Zulip):

I have a PR up that does so

pnkfelix (Oct 01 2020 at 14:17, on Zulip):

simulacrum said:

yep, and backport already happened actually so this is good

ah okay thanks @simulacrum

pnkfelix (Oct 01 2020 at 14:17, on Zulip):

(see #77308 for aforementioned backport PR, which explicitly notes that two of the patches were not yet approved...)

Santiago Pastorino (Oct 01 2020 at 14:18, on Zulip):

simulacrum said:

yep, and backport already happened actually so this is good

then we can close #76479

pnkfelix (Oct 01 2020 at 14:18, on Zulip):

hmm two ... but we only had one PR that we looked at above ...

pnkfelix (Oct 01 2020 at 14:18, on Zulip):

/me looks at the list of backported PR's on #77308

pnkfelix (Oct 01 2020 at 14:18, on Zulip):

oh did we approve #77063 last week or something?

Wesley Wiser (Oct 01 2020 at 14:19, on Zulip):

I thought we did

pnkfelix (Oct 01 2020 at 14:19, on Zulip):

yes we did. okay sorry for the confusion.

pnkfelix (Oct 01 2020 at 14:20, on Zulip):

Wesley Wiser said:

I have a PR up that does so

thanks for this @Wesley Wiser ! Okay so #77359 is also under control then.

pnkfelix (Oct 01 2020 at 14:20, on Zulip):
cuviper (Oct 01 2020 at 14:22, on Zulip):

Not directly a duplicate, just similar enough that they might have the same root cause

pnkfelix (Oct 01 2020 at 14:22, on Zulip):

is s390x also a big-endian 64-bit system?

cuviper (Oct 01 2020 at 14:22, on Zulip):

Yes

pnkfelix (Oct 01 2020 at 14:22, on Zulip):

I wonder if ARM64 is also big-endian.

cuviper (Oct 01 2020 at 14:23, on Zulip):

ISTR that arm can choose, but it's LE in our targets

pnkfelix (Oct 01 2020 at 14:24, on Zulip):

@ecstatic-morse made the excellent point on the comments for #77382 that if this is the same root cause, then #77382 provides us with a better starting point for investigation, since it is failing in a crate test case rather than during Firefox build process.

pnkfelix (Oct 01 2020 at 14:24, on Zulip):

but of course in the end we still have the problem of trying to reproduce bug for Tier2 platform that not all of us have immediate access to...

cuviper (Oct 01 2020 at 14:25, on Zulip):

I'm going to try and reduce the test case today, and start looking MIR / LLVM IR

pnkfelix (Oct 01 2020 at 14:26, on Zulip):

You may also want to look at whether applying PR #76081 fixes it for you.

oli (Oct 01 2020 at 14:26, on Zulip):

you don't even need to apply the PR, you can try just using -Zmir-opt-level=0

pnkfelix (Oct 01 2020 at 14:26, on Zulip):

(I know that we want to avoid landing PR #76081 since it does regress things, but maybe you can use it to provide more guidance into what is going wrong in #77382. Who knows, maybe there was a latent bug in PR #68528 ...)

pnkfelix (Oct 01 2020 at 14:27, on Zulip):

@oli but that disables even more mir-opts, right?

oli (Oct 01 2020 at 14:27, on Zulip):

yes

pnkfelix (Oct 01 2020 at 14:27, on Zulip):

which may make it more difficult to isolate the relevant change to the generated LLVM IR

oli (Oct 01 2020 at 14:27, on Zulip):

that's true

pnkfelix (Oct 01 2020 at 14:28, on Zulip):

is there a way to toggle individual mir-opts on/off via -Z flags?

pnkfelix (Oct 01 2020 at 14:28, on Zulip):

or actually, don't we have optimization fuel?

pnkfelix (Oct 01 2020 at 14:28, on Zulip):

/me looks

Wesley Wiser (Oct 01 2020 at 14:29, on Zulip):

Most mir-opts don't use optimization fuel

pnkfelix (Oct 01 2020 at 14:29, on Zulip):

oh. shouldn't they?

Wesley Wiser (Oct 01 2020 at 14:29, on Zulip):

I think MIR inlining is the only one that does

Wesley Wiser (Oct 01 2020 at 14:29, on Zulip):

Probably yeah :)

pnkfelix (Oct 01 2020 at 14:29, on Zulip):

Okay.

oli (Oct 01 2020 at 14:29, on Zulip):

also a way to individually enable/disable opts sounds like a good thing, too. I'm opening issues

oli (Oct 01 2020 at 14:30, on Zulip):

anyway, this is off topic for the actual thing we were talking about

pnkfelix (Oct 01 2020 at 14:30, on Zulip):

Opt fuel is definitely something I've found useful in other compiler projects for tracking down nice demo's of why somethings wrong

pnkfelix (Oct 01 2020 at 14:30, on Zulip):

Right

pnkfelix (Oct 01 2020 at 14:30, on Zulip):

For the actual thing we're talking about, #77382

pnkfelix (Oct 01 2020 at 14:30, on Zulip):

its currently marked P-critical

pnkfelix (Oct 01 2020 at 14:30, on Zulip):

this is a Tier2 platform

pnkfelix (Oct 01 2020 at 14:31, on Zulip):

but the number of different targets that might be getting hit by the same root bug is pretty scary

pnkfelix (Oct 01 2020 at 14:31, on Zulip):

I'm okay with leaving this P-critical for now

pnkfelix (Oct 01 2020 at 14:31, on Zulip):

but it should have someone assigned...

pnkfelix (Oct 01 2020 at 14:32, on Zulip):

@cuviper do you mind if I assign you, since it sounds like you're already actively investigating?

cuviper (Oct 01 2020 at 14:33, on Zulip):

That's fine

pnkfelix (Oct 01 2020 at 14:33, on Zulip):

great

pnkfelix (Oct 01 2020 at 14:33, on Zulip):

okay lets move along then

pnkfelix (Oct 01 2020 at 14:33, on Zulip):

libs-impl

T-rustdoc

pnkfelix (Oct 01 2020 at 14:33, on Zulip):

P-high regressions

P-high beta regressions

pnkfelix (Oct 01 2020 at 14:33, on Zulip):
pnkfelix (Oct 01 2020 at 14:35, on Zulip):

this does have an MCVE now, by the way

pnkfelix (Oct 01 2020 at 14:35, on Zulip):

https://github.com/rust-lang/rust/issues/76980#issuecomment-698473949

Santiago Pastorino (Oct 01 2020 at 14:36, on Zulip):

this is going to hit stable

pnkfelix (Oct 01 2020 at 14:37, on Zulip):

so the main issue is that this code is returning an empty string, according to a debugger investigation

pnkfelix (Oct 01 2020 at 14:37, on Zulip):

and that breaks invariants of some other code

pnkfelix (Oct 01 2020 at 14:37, on Zulip):

definitely seems goofy

pnkfelix (Oct 01 2020 at 14:38, on Zulip):

but it also isn't clear if we can do anything about it, based on the claims in the bug report.

Joshua Nelson (Oct 01 2020 at 14:38, on Zulip):

(deleted)

pnkfelix (Oct 01 2020 at 14:38, on Zulip):

I recently got rr working on my AMD box. I'd like to step backwards to find out why that code is returning an empty string, since I would think that impossible here.

Joshua Nelson (Oct 01 2020 at 14:39, on Zulip):

Reading the issue it seems like a bug in the library they're using?

Joshua Nelson (Oct 01 2020 at 14:39, on Zulip):

As in, the question is how it ever worked, not why it broke

pnkfelix (Oct 01 2020 at 14:39, on Zulip):

well I'd like to know if that os.str() call is allowed to return empty string

pnkfelix (Oct 01 2020 at 14:39, on Zulip):

if it is, then sure, that's definitely a bug

pnkfelix (Oct 01 2020 at 14:39, on Zulip):

anyway

pnkfelix (Oct 01 2020 at 14:39, on Zulip):

we're in the weeds

pnkfelix (Oct 01 2020 at 14:40, on Zulip):

I don't think we can do anything about this. and I don't think we're going to block the release based on it.

pnkfelix (Oct 01 2020 at 14:41, on Zulip):

lets move along. I'm not going to assign myself; I'm more vaguely hoping I remember to look at it later.

pnkfelix (Oct 01 2020 at 14:41, on Zulip):

Unassigned P-high nightly regressions

pnkfelix (Oct 01 2020 at 14:41, on Zulip):
pnkfelix (Oct 01 2020 at 14:41, on Zulip):

oh this was a great chat from last week

pnkfelix (Oct 01 2020 at 14:42, on Zulip):

and I think @eddyb had some theories on things we could do on our end to mitigate this

eddyb (Oct 01 2020 at 14:42, on Zulip):

oh nothing happened since

eddyb (Oct 01 2020 at 14:42, on Zulip):

yeah there's two solutions: pass a flag to revert behavior, or just do something universal

eddyb (Oct 01 2020 at 14:42, on Zulip):

(universal as in windows and non-windows would both parse it correctly)

pnkfelix (Oct 01 2020 at 14:42, on Zulip):

last weeks' discussion here

pnkfelix (Oct 01 2020 at 14:43, on Zulip):

It seems to me like we should first land a patch to pass the flag

pnkfelix (Oct 01 2020 at 14:43, on Zulip):

because that sounds least risky and also relatively little effort

pnkfelix (Oct 01 2020 at 14:43, on Zulip):

and then maybe follow up with the universal solution if that has benefits

pnkfelix (Oct 01 2020 at 14:44, on Zulip):

(in terms of having a simpler code base to maintain, uniformity across targets, etc)

eddyb (Oct 01 2020 at 14:44, on Zulip):

it's just removing a few lines I think

eddyb (Oct 01 2020 at 14:44, on Zulip):

the windows escaping should "just work" on non-windows

pnkfelix (Oct 01 2020 at 14:44, on Zulip):

i dunno, famous last words?

pnkfelix (Oct 01 2020 at 14:44, on Zulip):

anyway

pnkfelix (Oct 01 2020 at 14:44, on Zulip):

it would be good to fix this, but its not mission critical right now, right?

cuviper (Oct 01 2020 at 14:45, on Zulip):

If this will be a late beta backport, better to not affect other targets

pnkfelix (Oct 01 2020 at 14:45, on Zulip):

i.e. its only affecting nightly, not beta? (Though why doesn't it affect beta too? I thought beta had the LLVM 11 changes ...)

pnkfelix (Oct 01 2020 at 14:45, on Zulip):

/me double checks the claim that its nightly only

eddyb (Oct 01 2020 at 14:46, on Zulip):

the risk with passing the flag is older rust-lld not recognizing it, or maybe some users have custom setups

eddyb (Oct 01 2020 at 14:46, on Zulip):

but it seems unlikely to be a problem

eddyb (Oct 01 2020 at 14:46, on Zulip):

(if everyone uses wasm through our distributed rust-lld)

pnkfelix (Oct 01 2020 at 14:46, on Zulip):

am I complete misremembering where the LLVM upgrade is?

pnkfelix (Oct 01 2020 at 14:47, on Zulip):

its in beta now too right? Or is it about to be promoted to beta?

cuviper (Oct 01 2020 at 14:47, on Zulip):

It is in beta, yes

pnkfelix (Oct 01 2020 at 14:47, on Zulip):

so why doesn't #76466 also regress beta?

eddyb (Oct 01 2020 at 14:47, on Zulip):

it... should, right?

pnkfelix (Oct 01 2020 at 14:47, on Zulip):

is wasm a nightly-only feature?

pnkfelix (Oct 01 2020 at 14:47, on Zulip):

we need to figure this out

eddyb (Oct 01 2020 at 14:47, on Zulip):

also the escaping thing is almost hilarious in that the comment points out we could be using quotes but we don't for simplicity (and instead escape spaces)

pnkfelix (Oct 01 2020 at 14:47, on Zulip):

I'll assign myself this ticket

eddyb (Oct 01 2020 at 14:48, on Zulip):

but there's no actual complexity difference, it's just unnecessarily different

pnkfelix (Oct 01 2020 at 14:48, on Zulip):

to figure out if this is affecting beta too. Amazing we didn't cover that in our big discussion a week ago.

cuviper (Oct 01 2020 at 14:49, on Zulip):

It was bisected to 1.47-nightly exactly where LLVM 11 landed. That's definitely beta now

pnkfelix (Oct 01 2020 at 14:49, on Zulip):

Performance logs

Most significant changes this week came in response to regressions discussed in
last week's triage report. Curious readers may be interested in
#77058, in which the removal
of a single field from a struct caused a 25% decrease in wall-times for one
seemingly unrelated benchmark, or
#76986, an ABI change that
should be a pretty clear win but seems to have mixed results.

pnkfelix (Oct 01 2020 at 14:49, on Zulip):

(I'm going to move along, hopefully I will actually follow through on my above promise to investigate and then fix if needed.)

pnkfelix (Oct 01 2020 at 14:49, on Zulip):

Triage done by @ecstaticmorse.
Revision range: b01326ab033e41986d4a5c8b96ce4f40f3b38e30..6369a98ebdee8ce01510f5d4307ddb771c8cb0e5

0 Regressions, 1 Improvements, 3 Mixed

0 of them in rollups

pnkfelix (Oct 01 2020 at 14:49, on Zulip):

Improvements

#77041 perf: move cold path of process_obligations into a separate function

pnkfelix (Oct 01 2020 at 14:50, on Zulip):

Mixed

#77006 Cache eval_to_allocation_raw on disk

#76913 Fixing the performance regression of #76244

#76986 Return values up to 128 bits in registers

pnkfelix (Oct 01 2020 at 14:51, on Zulip):

Nags requiring follow up

pnkfelix (Oct 01 2020 at 14:51, on Zulip):

who does the nags?

simulacrum (Oct 01 2020 at 14:51, on Zulip):

(perf triagers)

pnkfelix (Oct 01 2020 at 14:51, on Zulip):

(or does mentioning it in this meeting act as the nag)

pnkfelix (Oct 01 2020 at 14:51, on Zulip):

okay.

pnkfelix (Oct 01 2020 at 14:52, on Zulip):

Nominated Issues

T-compiler

pnkfelix (Oct 01 2020 at 14:52, on Zulip):
simulacrum (Oct 01 2020 at 14:52, on Zulip):

We've since reverted const typeid on beta/master, so this is not too urgent anymore.

eddyb (Oct 01 2020 at 14:52, on Zulip):

const fn TypeId::of stabilization to be clear

pnkfelix (Oct 01 2020 at 14:52, on Zulip):

okay

pnkfelix (Oct 01 2020 at 14:53, on Zulip):

is there anything to try to discuss here? Or is this worth of an MCP and/or design meeting?

eddyb (Oct 01 2020 at 14:54, on Zulip):

unsure. we should maybe try to push "users" of TypeId's internal representation off of it

pnkfelix (Oct 01 2020 at 14:54, on Zulip):

hmm

eddyb (Oct 01 2020 at 14:54, on Zulip):

there are alternatives, like going through the Hash interface, that are effectively supported

pnkfelix (Oct 01 2020 at 14:54, on Zulip):

Maybe a future-incompat warning to people who observe it?

pnkfelix (Oct 01 2020 at 14:55, on Zulip):

(like passing it to transmute, I guess)

pnkfelix (Oct 01 2020 at 14:55, on Zulip):

okay well lets see what the other nominations are

eddyb (Oct 01 2020 at 14:55, on Zulip):

it might just be firefox that we care about, given the results of the crater run

pnkfelix (Oct 01 2020 at 14:55, on Zulip):

we can probably do something about that

pnkfelix (Oct 01 2020 at 14:55, on Zulip):

I hope

pnkfelix (Oct 01 2020 at 14:55, on Zulip):
pnkfelix (Oct 01 2020 at 14:56, on Zulip):

there's check boxes here, right?

Pietro Albini (Oct 01 2020 at 14:56, on Zulip):

I think we forgot to remove the I-nominated label

pnkfelix (Oct 01 2020 at 14:56, on Zulip):

ah they're all checked, excpet for niko and me

pnkfelix (Oct 01 2020 at 14:56, on Zulip):

great

pnkfelix (Oct 01 2020 at 14:56, on Zulip):
pnkfelix (Oct 01 2020 at 14:57, on Zulip):

So at this point both @Aaron Hill and @Vadim Petrochenkov have weighed in on that ticket saying that this was unambiguously a bugfix

pnkfelix (Oct 01 2020 at 14:58, on Zulip):

and that a warning-cycle is not feasible

pnkfelix (Oct 01 2020 at 14:58, on Zulip):

The main suggestion is that we doa "public announcement' warning about the change

pnkfelix (Oct 01 2020 at 14:58, on Zulip):

the relevant PR is tagged with relnotes

pnkfelix (Oct 01 2020 at 14:59, on Zulip):

namely #73084

pnkfelix (Oct 01 2020 at 14:59, on Zulip):

I'm now curous

pnkfelix (Oct 01 2020 at 14:59, on Zulip):

the relnotes, those go out with changes associated with the stable release, right?

pnkfelix (Oct 01 2020 at 15:00, on Zulip):

@simulacrum ^

pnkfelix (Oct 01 2020 at 15:00, on Zulip):

Perhaps part of the problem here is that firefox uses more recent builds than stable, and so they might get blindsided more easily?

simulacrum (Oct 01 2020 at 15:00, on Zulip):

yes

pnkfelix (Oct 01 2020 at 15:00, on Zulip):

that's something to think about, at least.

pnkfelix (Oct 01 2020 at 15:00, on Zulip):

anyway this seems like its handled, in the sense that we will have relnotes for this

pnkfelix (Oct 01 2020 at 15:01, on Zulip):

the only issue I can see is whether we should have beta/nightly relnotes?

pnkfelix (Oct 01 2020 at 15:01, on Zulip):

(nightly relnotes is probably silly)

nikomatsakis (Oct 01 2020 at 15:01, on Zulip):

beta relnotes seems plausible

pnkfelix (Oct 01 2020 at 15:01, on Zulip):

but: Our hour is up! and, huzzah, we actually got through the whole agenda this week??? Is it possible?

pnkfelix (Oct 01 2020 at 15:01, on Zulip):

incroyable

simulacrum (Oct 01 2020 at 15:01, on Zulip):

we usually cannot prepare relnotes fast enough for beta

nikomatsakis (Oct 01 2020 at 15:02, on Zulip):

another idea:

nikomatsakis (Oct 01 2020 at 15:02, on Zulip):

we could have a notification group

nikomatsakis (Oct 01 2020 at 15:02, on Zulip):

"relnotes"

nikomatsakis (Oct 01 2020 at 15:02, on Zulip):

and we just do @rustbot ping relnotes

pnkfelix (Oct 01 2020 at 15:02, on Zulip):

yeah there's some argument to be made that "if you're on beta, you need to be keeping track of the deltas"

nikomatsakis (Oct 01 2020 at 15:02, on Zulip):

and that tags it as relnotes, and pings folks who asked to be pinged

nikomatsakis (Oct 01 2020 at 15:02, on Zulip):

and we suggest that consumers like FF follow that

pnkfelix (Oct 01 2020 at 15:03, on Zulip):

this is a good point too.

pnkfelix (Oct 01 2020 at 15:03, on Zulip):

@nikomatsakis the heart of what you're saying, if I understand right,

pnkfelix (Oct 01 2020 at 15:03, on Zulip):

is that we don't need full release notes for beta

pnkfelix (Oct 01 2020 at 15:03, on Zulip):

just a summary list of the issues

pnkfelix (Oct 01 2020 at 15:03, on Zulip):

right?

pnkfelix (Oct 01 2020 at 15:03, on Zulip):

so if the relevant PR's are already appropriate tagged, that might suffice for beta customers, right?

nikomatsakis (Oct 01 2020 at 15:03, on Zulip):

well

nikomatsakis (Oct 01 2020 at 15:04, on Zulip):

that was different than what I was saying, but also seems true :)

pnkfelix (Oct 01 2020 at 15:04, on Zulip):

(this is arguably something that someone who knows enough github search queries should be able to do as a query)

pnkfelix (Oct 01 2020 at 15:05, on Zulip):

@nikomatsakis okay. I wasn't sure what pinging relnotes would do besides link people to those issues, which the relnotes label also does

pnkfelix (Oct 01 2020 at 15:05, on Zulip):

anyway it sounds like we should be able to address this reasonably, as long we don't set the bar so high as "have full documented relnotes for beta release"

pnkfelix (Oct 01 2020 at 15:05, on Zulip):

okay that's all folks, thanks for attending everyone in @T-compiler/meeting !

Santiago Pastorino (Oct 01 2020 at 15:08, on Zulip):

we will be filling next week agenda here

Santiago Pastorino (Oct 01 2020 at 15:09, on Zulip):

and we will have a checkin from @WG-traits

Santiago Pastorino (Oct 01 2020 at 15:09, on Zulip):

if the group has something that wants to share, feel free to fill it in the corresponding part of the agenda cc @nikomatsakis

nagisa (Oct 02 2020 at 00:31, on Zulip):

pnkfelix said:

I wonder if ARM64 is also big-endian.

cuviper said:

ISTR that arm can choose, but it's LE in our targets

not a super useful distinction as while the architecture may support this, almost nothing else does in practice.

cuviper (Oct 02 2020 at 00:43, on Zulip):

I double-checked, AArch64TargetMachine does go both ways

cuviper (Oct 02 2020 at 00:47, on Zulip):

Linaro also publishes aarch64_be cross-gcc toolchains

cuviper (Oct 02 2020 at 00:47, on Zulip):

:shrug:

nagisa (Oct 02 2020 at 01:24, on Zulip):

yeah, but what's gonna run them? kernels are predominantly le, userspace is predominantly built for le as well (and its unknown what can of worms it'll open once you start getting those things running on be)

Lokathor (Oct 02 2020 at 06:19, on Zulip):

yeah, ARM can switch endian in terms of chip design, but some chips are fused to a specific endian (usually little endian), and of course most software doesn't support switching so in practice that's not a big concern.

Last update: Nov 25 2020 at 02:30UTC