Stream: t-compiler

Topic: #53737 compiletest extern flag


qmx (Aug 29 2018 at 14:25, on Zulip):

@nikomatsakis any thoughts on @eddyb 's last comment on the issue?

nikomatsakis (Aug 29 2018 at 15:00, on Zulip):

didn't look at it yet, pointer?

qmx (Aug 29 2018 at 15:38, on Zulip):

https://github.com/rust-lang/rust/issues/53737#issuecomment-416640351

qmx (Sep 07 2018 at 12:49, on Zulip):

@nikomatsakis initial PR is up for review - curious to hear your thoughts on this one

nikomatsakis (Sep 07 2018 at 13:06, on Zulip):

cool, left a few comments

qmx (Sep 10 2018 at 13:41, on Zulip):

@nikomatsakis currently blocked at this question https://github.com/rust-lang/rust/pull/54020#issuecomment-419610870

nikomatsakis (Sep 10 2018 at 13:47, on Zulip):

@qmx you mean this?

is there a standard call for doing naming normalization (like the ones that make the output for a some-crate.rs file that emits a rlib to be called libsome_crate.so)?

If so .. uh .. I don't know =)

nikomatsakis (Sep 10 2018 at 13:47, on Zulip):

not that I know of

qmx (Sep 10 2018 at 13:47, on Zulip):

yep

nikomatsakis (Sep 10 2018 at 13:47, on Zulip):

do we have to have an extension?

qmx (Sep 10 2018 at 13:48, on Zulip):

tried without and it didn't work

nikomatsakis (Sep 10 2018 at 13:48, on Zulip):

that is, can we have the compiler hard-code it to be xxx or something using -o, or does that make windows fail or something

nikomatsakis (Sep 10 2018 at 13:48, on Zulip):

ok

nikomatsakis (Sep 10 2018 at 13:48, on Zulip):

probably you should just have write that fn yourself (sigh)

qmx (Sep 10 2018 at 13:48, on Zulip):

but yea, my worry is breaking windows

nikomatsakis (Sep 10 2018 at 13:48, on Zulip):

we could ping @Alex Crichton , though I they they are not monitoring Zulip, but they may know the answer

qmx (Sep 10 2018 at 13:49, on Zulip):

yea, looks like I'm a rabbit hole magnet :joy:

qmx (Sep 10 2018 at 14:03, on Zulip):

on a more serious note, if we've had a function that did exactly this (naming conventions), where would this thing live?

nikomatsakis (Sep 10 2018 at 14:11, on Zulip):

@qmx uh... it doesn't feel to me that there is an obvious place

nikomatsakis (Sep 10 2018 at 14:12, on Zulip):

I guess I could imagine making a rust-lang-nursery crate

nikomatsakis (Sep 10 2018 at 14:12, on Zulip):

that both rustc and compiletest depend on

nikomatsakis (Sep 10 2018 at 14:12, on Zulip):

I think that is what I would ideally want to see

nikomatsakis (Sep 10 2018 at 14:12, on Zulip):

it doesn't, I suppose, have to be a rust-lang-nursery crate

qmx (Sep 10 2018 at 14:29, on Zulip):

I'll start by trying to find where this is defined in the compiler - is driver a good guess as where to look first?

nikomatsakis (Sep 10 2018 at 15:04, on Zulip):

@qmx hmm well maybe ..

nikomatsakis (Sep 10 2018 at 15:04, on Zulip):

I think that OutputFilenames might be worth looking at

nikomatsakis (Sep 10 2018 at 15:04, on Zulip):

https://doc.rust-lang.org/nightly/nightly-rustc/rustc/session/config/struct.OutputFilenames.html

qmx (Sep 10 2018 at 15:05, on Zulip):

good, will take a look

nikomatsakis (Sep 12 2018 at 17:34, on Zulip):

@qmx so what is the status of https://github.com/rust-lang/rust/pull/54020/?

qmx (Sep 12 2018 at 17:40, on Zulip):

trying to find a cross-platform solution for it

qmx (Sep 12 2018 at 17:41, on Zulip):

this is going to take way longer since it looks like a pretty big rabbit hole

qmx (Sep 12 2018 at 17:41, on Zulip):

(which is fine)

qmx (Sep 12 2018 at 17:41, on Zulip):

the way the PR is now it's not ready for landing

qmx (Sep 12 2018 at 17:42, on Zulip):

essentially this breaks when you add aliasing into the mix

qmx (Sep 12 2018 at 17:42, on Zulip):

since it breaks the expectations of paths and names

nikomatsakis (Sep 12 2018 at 17:43, on Zulip):

essentially this breaks when you add aliasing into the mix

aliasing?

qmx (Sep 12 2018 at 17:44, on Zulip):

when you alias a crate

qmx (Sep 12 2018 at 17:44, on Zulip):

--extern foo=bar

nikomatsakis (Sep 12 2018 at 17:45, on Zulip):

isn't bar there .. path?

nikomatsakis (Sep 12 2018 at 17:45, on Zulip):

oh, I uess you just mean

nikomatsakis (Sep 12 2018 at 17:45, on Zulip):

the name foo is not the name the crate uses?

qmx (Sep 12 2018 at 17:45, on Zulip):

paths will live under foo, but the compiler barf if you don't provide the full path to bar

nikomatsakis (Sep 12 2018 at 18:13, on Zulip):

does this mean it just doesn't work at all?

nikomatsakis (Sep 12 2018 at 18:13, on Zulip):

or only if the cratenames don't match?

qmx (Sep 12 2018 at 18:14, on Zulip):

if the cratenames don't match

qmx (Sep 12 2018 at 18:14, on Zulip):

that's where the cross-platform concern kicks in

qmx (Sep 12 2018 at 18:14, on Zulip):

I'm going to invest more work on fixing this this week

nikomatsakis (Sep 12 2018 at 18:28, on Zulip):

that's where the cross-platform concern kicks in

why is that specifically an issue across platforms?

qmx (Sep 12 2018 at 18:52, on Zulip):

cause it involves filesystem paths

qmx (Sep 12 2018 at 18:53, on Zulip):

basically

qmx (Sep 12 2018 at 18:53, on Zulip):

and I want to make sure the test does "the right thing(tm)" on all supported platforms

nikomatsakis (Sep 12 2018 at 19:06, on Zulip):

ok :)

qmx (Sep 13 2018 at 15:00, on Zulip):

do you feel that restructuring the test suite would be something more important than the upcoming canonical grammar discussion?

nikomatsakis (Sep 13 2018 at 16:12, on Zulip):

it's perhaps a nice "side project" that is more actionable

nikomatsakis (Sep 13 2018 at 16:13, on Zulip):

I'd love to see compiletest making progress

nikomatsakis (Sep 13 2018 at 16:13, on Zulip):

there are some folks who've done good work on it in the past too that might be game to help out, I forget exactly who

qmx (Sep 13 2018 at 16:13, on Zulip):

I'm asking cause I'm heads into it now

nikomatsakis (Sep 13 2018 at 16:14, on Zulip):

into testing, you mean?

nikomatsakis (Sep 13 2018 at 16:14, on Zulip):

yeah, it's a nice continuation

qmx (Sep 13 2018 at 16:14, on Zulip):

yup

nikomatsakis (Sep 13 2018 at 16:14, on Zulip):

in a way it has a similar character :)

nikomatsakis (Sep 13 2018 at 16:14, on Zulip):

somebody needs to draw up a proposal

nikomatsakis (Sep 13 2018 at 16:14, on Zulip):

shop it around

qmx (Sep 13 2018 at 16:14, on Zulip):

it does

nikomatsakis (Sep 13 2018 at 16:14, on Zulip):

it seems like there is a reasonable amount of consensus though?

nikomatsakis (Sep 13 2018 at 16:14, on Zulip):

I feel like the main thing would be to capture people's "wants"

nikomatsakis (Sep 13 2018 at 16:14, on Zulip):

and make sure they're reasonably addressed

nikomatsakis (Sep 13 2018 at 16:15, on Zulip):

(e.g., @Vadim Petrochenkov's concern that having to run run-pass tests takes too long)

qmx (Sep 13 2018 at 16:15, on Zulip):

I kinda want to have a "grammar" for all the // flags

nikomatsakis (Sep 13 2018 at 16:15, on Zulip):

yeah, so the other option — which also makes a lot of sense! — is to avoid the big questions

nikomatsakis (Sep 13 2018 at 16:15, on Zulip):

and instead improve compiletest itself in small ways to start

nikomatsakis (Sep 13 2018 at 16:15, on Zulip):

until you have a handle on it

nikomatsakis (Sep 13 2018 at 16:15, on Zulip):

just documenting would be awesome

qmx (Sep 13 2018 at 16:15, on Zulip):

the split between early props and late props is awkward

nikomatsakis (Sep 13 2018 at 16:15, on Zulip):

rustc-guide docs are incomplete

nikomatsakis (Sep 13 2018 at 16:16, on Zulip):

the whole revisions scheme is a bit awkward, though handy..

qmx (Sep 13 2018 at 16:16, on Zulip):

one frustrating thing for sure is how hard is to land anything

nikomatsakis (Sep 13 2018 at 16:16, on Zulip):

just in general, or in compiletest specifically?

qmx (Sep 13 2018 at 16:16, on Zulip):

I'm used to it, but always leave a bad aftertaste

qmx (Sep 13 2018 at 16:16, on Zulip):

in general

qmx (Sep 13 2018 at 16:17, on Zulip):

it's probably related to the scope of the project

qmx (Sep 13 2018 at 16:17, on Zulip):

in that retrospect, splitting out things to external crates would be a huge improvement

qmx (Sep 13 2018 at 16:18, on Zulip):

I know I progressed a lot on my compiletest changes by just being able to cargo watch -x test

nikomatsakis (Sep 13 2018 at 16:21, on Zulip):

in that retrospect, splitting out things to external crates would be a huge improvement

This is definitely one of my motivations.

Last update: Nov 21 2019 at 13:05UTC