Stream: t-compiler/meetings

Topic: [weekly meeting followup] 2020-09-11 #54818


pnkfelix (Sep 11 2020 at 14:06, on Zulip):

Hi @T-compiler/meeting

pnkfelix (Sep 11 2020 at 14:07, on Zulip):

here is where I'll have my previously mentioned "open discussion time"

pnkfelix (Sep 11 2020 at 14:09, on Zulip):

My most immediate goal is to finish going through yesterday's agenda items

pnkfelix (Sep 11 2020 at 14:09, on Zulip):

but given the informal nature, feel free to inject digressions or post links to other Zulip topics that you think would be of general interest. :)

pnkfelix (Sep 11 2020 at 14:10, on Zulip):

In the performance logs for the last week, there were some small compile-time regressions injected

pnkfelix (Sep 11 2020 at 14:10, on Zulip):

Performance logs

A few small compile-time regressions this week. The first was #70793, which added some specializations to the standard library in order to increase runtime performance. The second was #73996, which adds an option to the diagnostics code to print only the names of types and traits when they are unique instead of the whole path. The third was #75200, which refactored part of BTreeMap to avoid aliasing mutable references.

pnkfelix (Sep 11 2020 at 14:10, on Zulip):

Triage done by @ecstatic-morse
Revision range: d927e5a655809b6b20501889e093c085d6ffe6f7..35fc8359868e65a0970094f648ba87fce634e8c7

3 Regressions, 0 Improvements, 0 of them in rollups.

pnkfelix (Sep 11 2020 at 14:10, on Zulip):

Regressions

#75200

#73996

#70793

pnkfelix (Sep 11 2020 at 14:11, on Zulip):

Improvements

pnkfelix (Sep 11 2020 at 14:11, on Zulip):

Nags requiring follow up

pnkfelix (Sep 11 2020 at 14:11, on Zulip):

let me see what that nag was about.

Joshua Nelson (Sep 11 2020 at 14:11, on Zulip):

pnkfelix said:

let me see what that nag was about.

https://rust-lang.zulipchat.com/#narrow/stream/247081-t-compiler.2Fperformance/topic/Triage.20nags.20and.20.2375200

Joshua Nelson (Sep 11 2020 at 14:12, on Zulip):

@ssomers is not very happy with the review process

pnkfelix (Sep 11 2020 at 14:15, on Zulip):

so I posted some followup thoughts in #t-compiler/performance > Triage nags and #75200

simulacrum (Sep 11 2020 at 14:16, on Zulip):

I don't think this nag needs too much follow-up

pnkfelix (Sep 11 2020 at 14:16, on Zulip):

nor do I

simulacrum (Sep 11 2020 at 14:16, on Zulip):

the btreemap code seems to be just very unstable

simulacrum (Sep 11 2020 at 14:17, on Zulip):

in the "slight adjustment" produces relatively strong effect. My current theory is it's very sensitive to inlining decisions or similar.

pnkfelix (Sep 11 2020 at 14:17, on Zulip):

gotcha

pnkfelix (Sep 11 2020 at 14:18, on Zulip):

/me is now skimming over the other metrics captures by perf here besides instruction counts

pnkfelix (Sep 11 2020 at 14:19, on Zulip):

the max-rss and the #faults both went up quite a bit on a few benchmarks

simulacrum (Sep 11 2020 at 14:20, on Zulip):

max-rss is really unstable

pnkfelix (Sep 11 2020 at 14:20, on Zulip):

but I think all of the significantly affected benchmarks are not representative of "real code" in the wild

pnkfelix (Sep 11 2020 at 14:20, on Zulip):

hmm

pnkfelix (Sep 11 2020 at 14:20, on Zulip):

@simulacrum in general, I assume you mean?

simulacrum (Sep 11 2020 at 14:20, on Zulip):

yes

simulacrum (Sep 11 2020 at 14:21, on Zulip):

we see variability of +/- 20% easily on comment changes

pnkfelix (Sep 11 2020 at 14:21, on Zulip):

I wonder if its worth consider changing the allocator just to make that number more stable

pnkfelix (Sep 11 2020 at 14:21, on Zulip):

tough to say

simulacrum (Sep 11 2020 at 14:21, on Zulip):

maybe, not sure. I've long wanted to figure out why there's such wide variability in the first place, I'd personally not expect it

pnkfelix (Sep 11 2020 at 14:22, on Zulip):

I've been on this ride before, in terms of the utility of specializing the benchmark host in such ways; doing so makes it diverge from the "real system", but if the measurements you're getting are too noisy to be useful ...

simulacrum (Sep 11 2020 at 14:23, on Zulip):

we've not historically worried too much about max-rss

pnkfelix (Sep 11 2020 at 14:23, on Zulip):

we probably should be more worried about it

simulacrum (Sep 11 2020 at 14:23, on Zulip):

if there's a consistent jump then maybe but usually it's not too interesting

simulacrum (Sep 11 2020 at 14:23, on Zulip):

yes, for sure

pnkfelix (Sep 11 2020 at 14:23, on Zulip):

in that we know that the memory use of the compiler makes development untenable for some people

pnkfelix (Sep 11 2020 at 14:24, on Zulip):

anyway, something to keep in mind for the future

pnkfelix (Sep 11 2020 at 14:24, on Zulip):

lets move along to the nominated issues for this week, see if there are any we can take care of even in an "unofficial" meeting like this

pnkfelix (Sep 11 2020 at 14:24, on Zulip):

Nominated Issues

T-compiler

pnkfelix (Sep 11 2020 at 14:24, on Zulip):
Wesley Wiser (Sep 11 2020 at 14:25, on Zulip):

I had added this item

simulacrum (Sep 11 2020 at 14:25, on Zulip):

I think we definitely don't have time for weekly triage

pnkfelix (Sep 11 2020 at 14:25, on Zulip):

in that we're already over-extended for each agenda each week?

simulacrum (Sep 11 2020 at 14:25, on Zulip):

right, yes

pnkfelix (Sep 11 2020 at 14:25, on Zulip):

its true

Wesley Wiser (Sep 11 2020 at 14:25, on Zulip):

Yeah, I agree.

pnkfelix (Sep 11 2020 at 14:25, on Zulip):

we could probably allocate one of teh Friday meeting slots to it though

Wesley Wiser (Sep 11 2020 at 14:26, on Zulip):

I think it would be valuable to maybe just to decide if this is something we even want to tackle.

pnkfelix (Sep 11 2020 at 14:26, on Zulip):

some of these predate the MCP process, right?

Wesley Wiser (Sep 11 2020 at 14:26, on Zulip):

Given the lack of activity on ~50% of these, arguably we could just close them as deferred.

pnkfelix (Sep 11 2020 at 14:26, on Zulip):

like, one step we could take is just review them and see how many could be turned into MCP'S

Wesley Wiser (Sep 11 2020 at 14:26, on Zulip):

I think nearly all of them do

Wesley Wiser (Sep 11 2020 at 14:27, on Zulip):

But some are cross-team and not necessarily appropriate for the MCP process (IMO)

pnkfelix (Sep 11 2020 at 14:27, on Zulip):

ah, sure

pnkfelix (Sep 11 2020 at 14:27, on Zulip):

those cases can stay as RFCs

pnkfelix (Sep 11 2020 at 14:27, on Zulip):

lets take a quick look now at the list, see if we can bucket them accordingly

Wesley Wiser (Sep 11 2020 at 14:28, on Zulip):

But before we spam a bunch of RFC authors with requests to use the MCP process, it might be good to decide if we even have interest in tackling these

pnkfelix (Sep 11 2020 at 14:28, on Zulip):

true

Wesley Wiser (Sep 11 2020 at 14:28, on Zulip):

I don't want to ask people to do work we're ultimately going to reject

pnkfelix (Sep 11 2020 at 14:28, on Zulip):

I'm going to try to review them here

simulacrum (Sep 11 2020 at 14:30, on Zulip):

are you thinking of inline commenting in that doc, or discussing one-by-one here?

pnkfelix (Sep 11 2020 at 14:32, on Zulip):

my plan is to first reorganize them

pnkfelix (Sep 11 2020 at 14:32, on Zulip):

ones with multiple team labels will go into a separate section

pnkfelix (Sep 11 2020 at 14:33, on Zulip):

the ones that are just tagged with us I feel a higher priority to address

pnkfelix (Sep 11 2020 at 14:34, on Zulip):

since they really are our sole responsibility to deal with

pnkfelix (Sep 11 2020 at 14:34, on Zulip):

so, now that I've bucketed them in the above manner

pnkfelix (Sep 11 2020 at 14:34, on Zulip):

I'm not sure it makes sense to go through them all one by one here

pnkfelix (Sep 11 2020 at 14:34, on Zulip):

but it may be possible to quickly identify things we can take action on

Wesley Wiser (Sep 11 2020 at 14:35, on Zulip):

I know I have thoughts about most of ours

Wesley Wiser (Sep 11 2020 at 14:35, on Zulip):

Should we maybe jot that stuff down in the doc and then see if there's any common feelings?

pnkfelix (Sep 11 2020 at 14:35, on Zulip):

sure, I'm happy to handle it that way

pnkfelix (Sep 11 2020 at 14:43, on Zulip):

so there's a lot of content to catch up on if one really wants to read each RFC before commenting on its content

pnkfelix (Sep 11 2020 at 14:44, on Zulip):

one detail I'm reflecting on, is that if you read the content of RFC PR #2400, there are some pretty strong claims made about whether what rustc currently does is valid in general

Wesley Wiser (Sep 11 2020 at 14:44, on Zulip):

I don't have any more comments to add about that set of RFCs

pnkfelix (Sep 11 2020 at 14:45, on Zulip):

and I think that its important to see that the comment feedback on RFC PR #2400 gives pretty good counter-arguments to those aforementioned claims

Wesley Wiser (Sep 11 2020 at 14:47, on Zulip):

I would fine with closing 2400 because it doesn't sound high priority for the compiler team at this time and it needs more design work.

pnkfelix (Sep 11 2020 at 14:47, on Zulip):

yes I think that is probably the right call

pnkfelix (Sep 11 2020 at 14:47, on Zulip):

actually, okay

pnkfelix (Sep 11 2020 at 14:47, on Zulip):

maybe now is the right time to just quickly go through the list here

pnkfelix (Sep 11 2020 at 14:47, on Zulip):
pnkfelix (Sep 11 2020 at 14:48, on Zulip):

So: Do you all think we should suggest to the author that they propose an MCP, or should we attempt to port the RFC to an MCP ourselves?

pnkfelix (Sep 11 2020 at 14:48, on Zulip):

I think the right answer is to ask the author to do it

simulacrum (Sep 11 2020 at 14:48, on Zulip):

Yes, I think so. In particular I would only want to accept this if someone's willing to drive the impl work as well

simulacrum (Sep 11 2020 at 14:48, on Zulip):

(no need to create debt)

pnkfelix (Sep 11 2020 at 14:48, on Zulip):

the main reason I balk at that is that I feel bad waiting all this time and then having the response be "oh sorry, use this other system we came up with in the meantime"

Wesley Wiser (Sep 11 2020 at 14:48, on Zulip):

The author doesn't appear to be very active on GitHub.

pnkfelix (Sep 11 2020 at 14:49, on Zulip):

but yes, @simulacrum 's point is crucial: We need someone willing to drive the work

pnkfelix (Sep 11 2020 at 14:49, on Zulip):

so the only RFC that I would really consider porting oursevles

Wesley Wiser (Sep 11 2020 at 14:49, on Zulip):

But it does sound like there are other interested parties who participated in the comments that might be willing to drive this forward.

pnkfelix (Sep 11 2020 at 14:49, on Zulip):

is one that we ourselves see as such a clear benefit that we ourselves would be willing to drive the work during our own time

simulacrum (Sep 11 2020 at 14:49, on Zulip):

I am okay fcp merging, but like, I don't want this to end up as an open issue that we end up needing to reconsider from scratch in a year because it's so far out of cache for everyone

pnkfelix (Sep 11 2020 at 14:50, on Zulip):

Yeah. Lets post a comment saying that we recommend this (RFC#2154) be converted to an MCP, either by the PR author or another interested volunteer.

pnkfelix (Sep 11 2020 at 14:50, on Zulip):

next

simulacrum (Sep 11 2020 at 14:50, on Zulip):

I can write that comment

pnkfelix (Sep 11 2020 at 14:51, on Zulip):
pnkfelix (Sep 11 2020 at 14:51, on Zulip):

as previously discussed here, I think we should just close this, as described by @Wesley Wiser

pnkfelix (Sep 11 2020 at 14:51, on Zulip):
pnkfelix (Sep 11 2020 at 14:51, on Zulip):

by "accept", that means propose an FCP merge?

Wesley Wiser (Sep 11 2020 at 14:52, on Zulip):

Yeah

pnkfelix (Sep 11 2020 at 14:52, on Zulip):

the MCP process may itself be lighter weight

Wesley Wiser (Sep 11 2020 at 14:52, on Zulip):

Or even potentially just merge it

pnkfelix (Sep 11 2020 at 14:52, on Zulip):

oh yes

simulacrum (Sep 11 2020 at 14:52, on Zulip):

well, I would be fine just merging it

pnkfelix (Sep 11 2020 at 14:52, on Zulip):

its just an amendment

Wesley Wiser (Sep 11 2020 at 14:52, on Zulip):

As niko said:

We often do an FCP for amendments, but I also think it's ok to "just do it" if it is small and not controversial amongst the folks who were active in the RFC itself. I am fine with this change (but I've not been following the RFC closely).

pnkfelix (Sep 11 2020 at 14:52, on Zulip):

okay yeah its just an amendment

pnkfelix (Sep 11 2020 at 14:52, on Zulip):

lets just merge it then

pnkfelix (Sep 11 2020 at 14:52, on Zulip):

next

pnkfelix (Sep 11 2020 at 14:52, on Zulip):
Wesley Wiser (Sep 11 2020 at 14:53, on Zulip):

Seems like a good candidate for MCP especially as the author is quite active in the project.

pnkfelix (Sep 11 2020 at 14:54, on Zulip):

here again I think we can take the same approach that was suggested for rfc#2154: suggest it be converted to MCP, either by PR author or by another interested contributor

pnkfelix (Sep 11 2020 at 14:54, on Zulip):
pnkfelix (Sep 11 2020 at 14:54, on Zulip):

so based on the comments here

simulacrum (Sep 11 2020 at 14:54, on Zulip):

(It was unclear to me why this was an RFC, to be honest)

pnkfelix (Sep 11 2020 at 14:55, on Zulip):

the main reason why this might not be "just" a PR

pnkfelix (Sep 11 2020 at 14:55, on Zulip):

is that the implementation is "hard"

pnkfelix (Sep 11 2020 at 14:55, on Zulip):

at least doing it well

Wesley Wiser (Sep 11 2020 at 14:55, on Zulip):

The author may have also been trying to get feedback from diagnostics people if this would be accepted or not before doing the work.

simulacrum (Sep 11 2020 at 14:55, on Zulip):

I think that's a good candidate for MCP today -- to get buy-in on idea before working on impl

pnkfelix (Sep 11 2020 at 14:56, on Zulip):

right, and then maybe it requires a project group, maybe not

simulacrum (Sep 11 2020 at 14:56, on Zulip):

but it also seems like the kind of thing where we usually have "just" a diagnostics issue on rust-lang/rust for

simulacrum (Sep 11 2020 at 14:56, on Zulip):

(or, half a dozen targeting different instances of this)

Joshua Nelson (Sep 11 2020 at 14:56, on Zulip):

pnkfelix said:

in that we know that the memory use of the compiler makes development untenable for some people

for context, I regularly run out of ram compiling docs.rs

Joshua Nelson (Sep 11 2020 at 14:57, on Zulip):

and just have to restart my whole laptop because of thrashing

pnkfelix (Sep 11 2020 at 14:57, on Zulip):

yeah back in the day @eddyb had the same problem, @Joshua Nelson

pnkfelix (Sep 11 2020 at 14:57, on Zulip):

it led them to invest substantial effort in getting rustc's peak memory usage down

pnkfelix (Sep 11 2020 at 14:58, on Zulip):

anyway, regarding rfc#2777

pnkfelix (Sep 11 2020 at 14:59, on Zulip):

lets write a comment suggesting that the RFC be converted to an MCP

pnkfelix (Sep 11 2020 at 14:59, on Zulip):

oh, we didn't actually say wehther we were going to close such RFCs

pnkfelix (Sep 11 2020 at 14:59, on Zulip):

I ... think the right thing would be a label

pnkfelix (Sep 11 2020 at 14:59, on Zulip):

something like "convert-to-mcp" or "should-be-mcp"

simulacrum (Sep 11 2020 at 15:00, on Zulip):

I would be inclined to close, personally

pnkfelix (Sep 11 2020 at 15:00, on Zulip):

and we leave the RFC open until an MCP is created, or until some amount of time has passed

simulacrum (Sep 11 2020 at 15:00, on Zulip):

But maybe, say, in a month

pnkfelix (Sep 11 2020 at 15:00, on Zulip):

yeah, lets give it a month (or maybe six weeks, heh)

Wesley Wiser (Sep 11 2020 at 15:01, on Zulip):

That seems good to me

pnkfelix (Sep 11 2020 at 15:01, on Zulip):

I just want to make sure that 1. good ideas don't get lost in the shuffle and 2. contributors don't get too discouraged , which I think closing can cause, even with really nice comments tacked on

pnkfelix (Sep 11 2020 at 15:01, on Zulip):

so okay, next

pnkfelix (Sep 11 2020 at 15:01, on Zulip):
pnkfelix (Sep 11 2020 at 15:02, on Zulip):

is the parselib WG also dealing with macros?

pnkfelix (Sep 11 2020 at 15:02, on Zulip):

oh I guess tat detail doesn't matter

pnkfelix (Sep 11 2020 at 15:02, on Zulip):

I mean, it does matter

pnkfelix (Sep 11 2020 at 15:02, on Zulip):

but the point is more that this RFC is clearly subsumed by the parselib WG, right?

Wesley Wiser (Sep 11 2020 at 15:03, on Zulip):

That's my feeling yeah

Wesley Wiser (Sep 11 2020 at 15:03, on Zulip):

Not that WGs can't submit RFCs but I feel like the project has already signed off on this work at least to some extent.

pnkfelix (Sep 11 2020 at 15:03, on Zulip):

lets close this and say that active work is going on with parselib WG, and link to the appropriate pages describing the WG and/or MCPs

pnkfelix (Sep 11 2020 at 15:04, on Zulip):

well, that, or

pnkfelix (Sep 11 2020 at 15:04, on Zulip):

parselib will come up with something and propose a new RFC

pnkfelix (Sep 11 2020 at 15:05, on Zulip):

but yeah, I think it would be weird to keep this open when I think the most important step has been taken

pnkfelix (Sep 11 2020 at 15:05, on Zulip):

in terms of telling these contributors "yes, we want something here; go figure out what to do."

pnkfelix (Sep 11 2020 at 15:05, on Zulip):

so okay, close this with links to places tracking ongoing work

pnkfelix (Sep 11 2020 at 15:06, on Zulip):
pnkfelix (Sep 11 2020 at 15:07, on Zulip):

I'm skimming over the RFC text now

pnkfelix (Sep 11 2020 at 15:08, on Zulip):

I don't currently understand exactly how the modifiers here translate into what changes to the linker invocation

pnkfelix (Sep 11 2020 at 15:09, on Zulip):

ah okay

pnkfelix (Sep 11 2020 at 15:09, on Zulip):

each modifier has a specific flag (or set of flags) it translates to, and that may be linker-dependant

pnkfelix (Sep 11 2020 at 15:10, on Zulip):

(and the spec for each modifier states its behavior on linkers that don't support it ... which I think in all given cases is "no op if unsupported", right?)

pnkfelix (Sep 11 2020 at 15:11, on Zulip):

the reason I'm digging into this level of detail is that I'm trying to figure out if this is indeed something that needs to loop in T-lang as well

pnkfelix (Sep 11 2020 at 15:12, on Zulip):

I think overall I'm in favor of moving forward with it unstably

pnkfelix (Sep 11 2020 at 15:12, on Zulip):

we can resolve the T-lang part concurrently

pnkfelix (Sep 11 2020 at 15:13, on Zulip):

i.e. the main thing I can imagine T-lang having an opinion on is whether its simply a bad idea to stuff this into into the #[link] attribute

pnkfelix (Sep 11 2020 at 15:13, on Zulip):

and that's a discussion that doesn't need to block unstable development of the feature

pnkfelix (Sep 11 2020 at 15:13, on Zulip):

lets go ahead and suggest it be posted as an MCP

pnkfelix (Sep 11 2020 at 15:14, on Zulip):

same as other cases above

pnkfelix (Sep 11 2020 at 15:14, on Zulip):

okay, we're well past the hour

pnkfelix (Sep 11 2020 at 15:14, on Zulip):

thanks to @simulacrum and @Wesley Wiser for helping out here

Wesley Wiser (Sep 11 2020 at 15:14, on Zulip):

Sorry for my topic dominating the hour but it feels like we made a huge amount of progress!

Wesley Wiser (Sep 11 2020 at 15:14, on Zulip):

I honestly didn't expect we'd get that far.

pnkfelix (Sep 11 2020 at 15:15, on Zulip):

(and to everyone in @T-compiler/meeting for attending)

pnkfelix (Sep 11 2020 at 15:15, on Zulip):

yeah this was a good use of time I think

simulacrum (Sep 11 2020 at 15:15, on Zulip):

@pnkfelix are you planning to follow-up and post these comments? I volunteered to do 2154 and can do 2256 as well if you'd like

pnkfelix (Sep 11 2020 at 15:16, on Zulip):

well

pnkfelix (Sep 11 2020 at 15:16, on Zulip):

my current plan is to go back through yesterday's meeting

pnkfelix (Sep 11 2020 at 15:16, on Zulip):

and make sure comments got posted where appropriate for that

pnkfelix (Sep 11 2020 at 15:16, on Zulip):

when I'm done with that, I'll do the same for this meeting

pnkfelix (Sep 11 2020 at 15:16, on Zulip):

anyone who wants to take one of the RFC's listed above, feel free. :)

pnkfelix (Sep 11 2020 at 15:17, on Zulip):

(I suppose it would be good to use some common template for the common pattern of "port this to MCP")

simulacrum (Sep 11 2020 at 15:19, on Zulip):

okay. I'll take 2256 and 2154. Probably won't have a chance to type those up until later today, but there's no real rush I guess :)

pnkfelix (Sep 11 2020 at 15:22, on Zulip):

by the way, @simulacrum , do you indeed think you'll have a chance to make a PR for #74753 that disables the check locally?

simulacrum (Sep 11 2020 at 15:23, on Zulip):

yes it's on my todo list

pnkfelix (Sep 11 2020 at 15:23, on Zulip):

(and now I'm wondering what that meant? as in, the check will never run on people's local machines, just on CI alone?)

simulacrum (Sep 11 2020 at 15:23, on Zulip):

that is my plan yes

simulacrum (Sep 11 2020 at 15:23, on Zulip):

it is a bit unfortunate but ultimately CI is the only place I know of where we can assume things about the filesystem, and I don't want to try to write logic that detects filesystem properties

pnkfelix (Sep 11 2020 at 15:24, on Zulip):

hmm. Just because that's less fragile than this logic to disable it on certain hosts based on inspecting the /proc/version ?

simulacrum (Sep 11 2020 at 15:24, on Zulip):

/proc/version tells you nothing about the filesystem afaict, which is what you really want

pnkfelix (Sep 11 2020 at 15:24, on Zulip):

Right,

pnkfelix (Sep 11 2020 at 15:24, on Zulip):

yeah "fragile" was probably not the right word

pnkfelix (Sep 11 2020 at 15:25, on Zulip):

or at least, its not the only thing that concerns us about the logic here

pnkfelix (Sep 11 2020 at 15:25, on Zulip):

okay

pnkfelix (Sep 11 2020 at 15:25, on Zulip):

I guess thats' fine

pnkfelix (Sep 11 2020 at 15:26, on Zulip):

though I'm now musing about filesystem inspection

pnkfelix (Sep 11 2020 at 15:27, on Zulip):

wouldn't it suffice to touch a fresh file and then inspect its exec bit?

simulacrum (Sep 11 2020 at 15:27, on Zulip):

hm, maybe

simulacrum (Sep 11 2020 at 15:27, on Zulip):

but then I wonder ... why haven't we already done that?

pnkfelix (Sep 11 2020 at 15:27, on Zulip):

maybe because we don't want to write into the source directory

pnkfelix (Sep 11 2020 at 15:28, on Zulip):

its possible that someone might actually have it read-only

simulacrum (Sep 11 2020 at 15:28, on Zulip):

hm, maybe

simulacrum (Sep 11 2020 at 15:28, on Zulip):

(though arguably if so we can just say "well probably you're not editing it either")

pnkfelix (Sep 11 2020 at 15:28, on Zulip):

(and is doing the build elsewhere ... not that I've ever done a build in a directory that isn't the root directory nor a subdirectory of the root)

pnkfelix (Sep 11 2020 at 15:29, on Zulip):

sure, but the tidy run doesn't know whether you've tried to exit edit

pnkfelix (Sep 11 2020 at 15:29, on Zulip):

that would be funny, actually

simulacrum (Sep 11 2020 at 15:29, on Zulip):

I guess I added it to fix https://github.com/rust-lang/rust/issues/36706

pnkfelix (Sep 11 2020 at 15:29, on Zulip):

e.g. first do git status or somesuch, and then only if git status reports unclean, then touch a file and inpsect its exec bit

simulacrum (Sep 11 2020 at 15:29, on Zulip):

it = /proc/version check

simulacrum (Sep 11 2020 at 15:30, on Zulip):

I might try the touch a file strategy

pnkfelix (Sep 11 2020 at 15:30, on Zulip):

I have to admit, I would prefer some way to avoid a check living solely in CI

simulacrum (Sep 11 2020 at 15:31, on Zulip):

and we can just skip the bins check if it's not possible to touch a file

pnkfelix (Sep 11 2020 at 15:31, on Zulip):

:+1:

pnkfelix (Sep 11 2020 at 15:31, on Zulip):

I like that

pnkfelix (Sep 11 2020 at 15:31, on Zulip):

that seems very likely to cover the cases of interest

simulacrum (Sep 11 2020 at 15:31, on Zulip):

:thumbs_up:

pachi (Sep 11 2020 at 15:37, on Zulip):

@pnkfelix RFC 2951 has an MCP https://github.com/rust-lang/compiler-team/issues/356

pachi (Sep 11 2020 at 15:37, on Zulip):

MCP

pnkfelix (Sep 11 2020 at 15:38, on Zulip):

doh

pnkfelix (Sep 11 2020 at 15:38, on Zulip):

thanks @pachi

pnkfelix (Sep 11 2020 at 15:39, on Zulip):

hmm

pnkfelix (Sep 11 2020 at 15:39, on Zulip):

actually

pnkfelix (Sep 11 2020 at 15:39, on Zulip):

this is a public facing feature, isn't it

pachi (Sep 11 2020 at 15:39, on Zulip):

And I think libsyntax2.0 was some @matklad 's proposal that I feel should be now subsumed by some joint rustc / rust-analyzer librarifycation and the new parser work

pnkfelix (Sep 11 2020 at 15:41, on Zulip):

right, I think we reached the same conclusion in this discussion

pachi (Sep 11 2020 at 15:41, on Zulip):

Ok, sorry, I was just skimming through it

pnkfelix (Sep 11 2020 at 15:42, on Zulip):

hey @LeSeulArtichaut or @simulacrum , I don't know if I screwed something up over here: https://github.com/rust-lang/compiler-team/issues/356#issuecomment-691169524

pnkfelix (Sep 11 2020 at 15:43, on Zulip):

is there a way to undo a second so that I can make it a proper "fcp merge" instead?

pnkfelix (Sep 11 2020 at 15:43, on Zulip):

(do I need to close and reopen the issue or something?)

simulacrum (Sep 11 2020 at 15:43, on Zulip):

they're to different bots

pnkfelix (Sep 11 2020 at 15:43, on Zulip):

oh sorry

simulacrum (Sep 11 2020 at 15:43, on Zulip):

you should put rfcbot fcp merge

pnkfelix (Sep 11 2020 at 15:43, on Zulip):

I think rustbot links to triagebot's doc

pnkfelix (Sep 11 2020 at 15:43, on Zulip):

oh that's the other bot you mean

pnkfelix (Sep 11 2020 at 15:44, on Zulip):

hmm

pnkfelix (Sep 11 2020 at 15:44, on Zulip):

(maybe rustbot should say so when you ask it to fcp merge something)

LeSeulArtichaut (Sep 11 2020 at 15:44, on Zulip):

(I think you wrote rfc instead of fcp)

simulacrum (Sep 11 2020 at 15:44, on Zulip):

I think it should too. I would've expected an unknown command comment

pnkfelix (Sep 11 2020 at 15:44, on Zulip):

omg

pnkfelix (Sep 11 2020 at 15:45, on Zulip):

/me needs to learn to cut-and-paste

Joshua Nelson (Sep 11 2020 at 15:48, on Zulip):

are there docs for this somewhere?

LeSeulArtichaut (Sep 11 2020 at 15:49, on Zulip):

I don't think there is any doc about @rfcbot, probably should go in Forge

Joshua Nelson (Sep 11 2020 at 15:50, on Zulip):

https://github.com/rust-lang/rust-forge/issues/437

Joshua Nelson (Sep 12 2020 at 16:22, on Zulip):

does anyone mind if I add A-contributor-roadblock?

Joshua Nelson (Sep 12 2020 at 16:23, on Zulip):

like D-newcomer-roadblock but for people working on rustc itself

Wesley Wiser (Sep 13 2020 at 15:36, on Zulip):

Seems fine to me

Joshua Nelson (Sep 13 2020 at 16:18, on Zulip):

added, thanks

Last update: Nov 25 2020 at 01:30UTC