Rust Programming Language Rust and WebAssembly Working Group, 14 Mar 2019

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: 2019-03-14 Rust and WebAssembly Working Group Meeting

Description

https://github.com/rustwasm/team/issues/260

Happy Pi day :)

A

Welcome to another Rustin webassembly working group meeting thanks for coming everyone, so we're gonna skip our usual RC triage and status updates today, because we talked about last time we wanted to plan, or we wanted to resume six week Sprint's and wanted to plan out this.

A

This sprint I guess what our goals are going to be, but before we get to that, just like real, quick, just checking things off in case you didn't see it. There was the announcement for glue the modular toolkit that we planned in our roadmap for 2019, and so it doesn't actually really have anything yet, but this is kind of a call for participation. So if you want to get more involved, that's a great place to do it.

A

Another thing is that someone some higher made a list of NPM packages that they thought. Maybe there could possibly be a restaurant version, and so, if you are interested in doing a resin version of an NPM package, this list looks like a good place to start.

A

So the six-week, sprint planning so six weeks from now is April 25th. Just so we have an idea of the timeframe. I think we should add a sprint label to all the issues, and then that makes it really easy to view all the things in the sprint across the whole organization, which I found very useful.

A

Last time we did these things.

A

And yeah, so our kind of top level goals called out in our roadmap are the.

A

The glue tool kit.

A

Multi-Threading.

A

A

A

Monitoring stuff.

A

Everything out debugging.

A

So I guess we could start with the glue toolkit in six weeks hmm what we have so we're still very much in kind of the design and planning and figuring out what the heck we're doing with glue for six weeks. So I think it makes sense to say we should have a clear picture of what we're doing and like be.

A

In the middle of implementing I guess or like getting implementation off the ground.

A

Do you have any thoughts about that upon.

A

B

A

So I guess what can happen is I can tag more of these things as sprint issues, as we figure out we're doing.

A

Okay, multi-threading stuff.

B

um I, don't I, don't think. No, if there's a hole, I think maybe the goal could be. We should get the Ray on demo landed.

C

B

Something using the experimental branch of Ray on that Josh is that has been prepared for us.

B

And so yeah, that's probably the best extent which we can go here.

A

D

So I read the agenda and saw the plan about the issues stuff, so ODOT a thought Oh, which is the release that will come after tomorrow's release, is the same as sprint and so I just went through and all of the issues that were already marked there as Oda Eiichiro I marked his sprint as well, unless I already had a PR attached, which we have a couple, and so in general I did this based on kind of like where we are in the process, but like this doesn't mean that we can't change what I've put here, I just let having a straw man.

D

It would be good and so for features. I felt like getting wise impact generate and then well as an OP support and then a target for producing isomorphic apps. Those are three relatively big, desirable features that I think we should be able to accomplish in six weeks and then from there regarding support. This was largely because of pr's. We kind of already had open that some people have made kind of vague progress on which is support for non rust up set ups, an NPM, installer I.

D

Don't know if there's other support we wanted, but those were kind of just already in flight and then for maintenance. I. Think the real big one here is the improvements of CLI output strategy and then also there's kind of an ask to do improve panics and then also move binary installed into its own crate and then finally, for Docs mostly contribute facing the release checklist and then improved contributing guidelines. It's kind of what I was thinking but very open to feedback.

A

That sounds like a good collection of things.

A

Yeah, so the the longline stuff- that's is that fixed by your PR, Alex I believe so. Yes,.

D

Yeah I've been tested to see if it is, but my assumption was that it was.

A

Do we think we can get the.

A

Local, not the local Snippets, the external crate, depending on NPM packages, stuff Oh in this six weeks as well.

D

I have not looked exactly or like thought exactly about how we can implement that, but I think we should be able to. We should I should file an issue to make sure that we're tracking that I think that's why that missed my. But if that support is gonna, let in wise and bind John I want to support it and was in pack ASAP. So I would like prioritize that over maybe like a cargo generate just because it seems more important to me. I, don't know what you think about that, but.

A

D

Think we should probably be able to do both, but I want to be it's especially for, like the first sprint in this set of Sprint's it'd be nice to be conservative and successful.

D

A

And so the other thing to note about the sprint label was like I intentionally, didn't do like sprint to zero or sprint one, because it is there's no hard deadlines. Really it's just like goals, so they can just automatically roll over cool.

D

Yeah I'll create a tracking issue for doing that and I'll definitely label it sprint and I'll put it in a date, but we'll see maybe something will slip if we end up doing it but seems like a thing that we should get in quickly.

A

All right uh performance, slash, good science, monitoring, stuff I, found that the chrome puppeteer stuff has some interesting time to interaction measurements which, if there's a cross-browser solution based on webdriver I, would prefer using that. But if not I'm, okay with using the one thing that we do have available anyway. So I was thinking that it would be nice to maybe stand up some like time to interaction measurements next to the code size stuff, because this is kind of the thing faster escs that that we always have to add. Whenever we talk about code size.

A

Is that like this is the best proxy we currently have for measuring the actual thing? And the actual thing is more like time to interactive.

A

Does that sound like a good goal for for the six-week, sprint, I think.

D

So and I'm a big fan of the puppeteer project so I think leveraging. That would be good and if we want any help, former colleague of mine at NPM there's a lot of work on that I.

B

Think it'd be worthwhile to get a proof of concept. I'm a little wary in terms of performance monitoring is really really difficult to do on CI without dedicated hardware. So.

C

B

It might mean that it'll be really difficult to get useful information out of it like in terms of the link over time, but having something that you can run locally and we could verify at least ourselves would be a lot better than nothing. We have right now, yeah.

A

Or like maybe it's just making a little wrapper, that makes it easy to run this on rust, resin projects, and then we can just have it kind of separate from the weight tracker just as a tool that we can point people to when they start asking about code size. What's good boy, yeah.

A

All right and debugging stuff I like to have walrus tracking the origin of all of the new instructions and transformations and I guess existing instructions and not necessarily doing anything with it. Yet, but I think we should have that tracking their.

B

Spin asks and Chad what's the status of source maps or dwarf and wasm.

A

There's been little movement in the browser world I.

A

Think for many months in in kind of the crane lift world there's a proof of concept of transforming a dwarf that was emitted for the Wazza into a dwarf. That's for the native code that that crane left is emitting, and so you can debug that stuff with lltv, which is pretty cool.

A

A

All right any other things that we would like to call out for the sprint anyone's pet issue.

D

There's not necessarily a pet issue. This is just sprint in any category I guess so tomorrow, I'm gonna land. What is not a perfect but just needs to land version of this, like marketing page for magazine pack, with like kind of the funnels based on target that helps like a line like what you're targeting and like what template you. She is what workflow etc.

D

It would be great if we could try to push the like unification of like the landing page. Cuz like this is just gonna, be on like the home page of Waze impact right now, but this is supposed to be like a unified tool chain. So just having it be a page for as impact seems like a little silly I, don't know.

D

If there's anybody interested in that, but once that lands tomorrow, I think it would be nice to try and get it all in the same place and feel like more of a like funnel for the whole project. I.

A

Mean if you saw it but J finished, Alex's Docs, bookshelf, PR thing which is kind of the same thing, but on the dark side rather than the landing page side. So that's pretty sweet ya.

D

Know it's super awesome and I. Think that's why I like having like the docks and having those funnels and then kind of like the marketing like get started funnels all in the same place would be like super great.

A

Yeah, so the only other thing that I'd like to aim for is like it seems we have a lot of RFC's that are very near to completion, and if there's not the concerns and non-consensus, it would be pretty sweet to get those merged.

A

And then maybe in the next sprint we could make those RFC's things goals.

D

So I checked my box in the MPM depths dependency and I can check my box on.

D

On the roadmap, one, the other one that's open- is the RFC for, like my duplicate NPM, one that Alex scooped up in my career ode, the only hesitation I have about the RFC for watch and serve on wesson pack. I think we should definitely do it. I would like to approve and merge in RFC close to when it will be implemented, though, and I'm worried that it might be like a little while until it gets that, since we do have a bit of a backlog that I think we want to prioritize over those things.

D

So like is there a way we can say like put on hold to like? We think we're ready to do some implementation, because you know the interface of wesson pack we're adding a bunch of commands, it might change, and so like I, don't want to approve now and then have to like edit it a ton later.

A

I guess I was more optimistic about our ability to implement it quickly. Yeah.

D

I guess I'd want to be conservative, like maybe we get towards the end it like. We certainly are not going to implement it in the next sprint, so I think. As we approach the end of the next sprint, we can like evaluate whether or not we think it's something that we can land and then like approve it closer to like that sprint, so that we can then do that if that makes any sense.

A

Why are we certainly not going to influence it in the next print? Well,.

D

As an RFC, its design hasn't yet landed, and the next sprint starts today and we have.

C

A pretty cool set.

D

Of things that we want yeah I'm.

B

Not like I totally.

D

Think it's possible, but just like I don't know it seems like we just did the planning and it didn't come up. So this.

A

Was a I thought: I was thinking this sprint and next sprint. I'm. Sorry.

D

Yeah those words are very vague: I apologize, yeah, okay,.

A

Yeah I would be happy was saying we should I. I would like to target implementing those things, not this sprint next print, because.

C

A

To have the note, am/pm stuff in this tutorial and I think we need to have this in order to do that, and.

A

So I think it makes sense to to aim for having the RFC merge at the end of this coming sprint so that we can start on it for the next sprint. Okay,.

D

Awesome I'm extremely down with that. I think that that works with the timeline I'd expect.

A

This is, of course, assuming you know, fundamental concerns and lack of consensus. Yeah.

D

So far it's been very positive, so I think I think most people wanted. For me it was really just a big question of win.

A

All right so yeah after the meeting I'll go through all this stuff and anything. That's not already tagged sprint or doesn't already have an issue, a file, an issue and then, of course, you know we can adjust as as the week goes on.

A

A

A

Alex, you added a agenda item about handling unpaired, surrogates and JavaScript strings. Well,.

B

This is actually something brought up by Palin this morning, so let them take it away. I.

B

Just added a fancy heading, so just an issue. Thank you.

B

So I think that general idea is that there was a issue on Palin's Dominator library's showing the gia strings can actually have unpaired utf-16 circuits, and so you can't actually losslessly receive a string in rust right now, because we always do run it through utf-8, encoding and that'll actually replace these circuits with replacement characters, I believe, and so the issue is describing, though this is probably a bug and probably something about a fixed Muslim engine where, if you I kind of make sense that right now, if you have amber sensor what we have to utf-8 encoded.

B

But if we might want to have the option of like an oyster like thing where it's a lossless representation of what's in jeaious- and it says a fallible to utf-8 kind of thing or something like that. Isn't.

A

B

A

B

16 I'm, actually not a hundred percent. Certain I, don't know about J's strings I know for Windows. Those are just arbitrary, u16 sequences. Since that's what utf-8 is but I get the impression that Gaea strings are not arbitrary, u16 sequences, but rather arbitrary sequences of Unicode something's unpaired. Actually, no wait. I think.

A

Simone has a crate for dealing with these things like, like I, think he's also the standards person for WTF, eight and I think there's a sixteen variant. I think that's.

A

But like yeah, it certainly seems like like, like we don't, have any fundamental unsafety now, because we always in code right, but there is an express ability problem and that we can't express things that JavaScript can and in reflection, probably shouldn't but yeah. So we should have support for for these sorts of strings that exactly match the semantics. That JavaScript has.

A

Seems fairly uncontroversial to me, I.

B

Think there's also a question of how we want to handle this with a web IDL bindings, so do want to take every string type and explode it to either aster or the Oster like thing, but I think we can probably handle that. So you follow up step as opposed to handling them right now.

A

Yeah I think we can take this discussion to the issue.

B

Spin has some good links to some web assembly store uh yeah to some other related issues.

A

So what is this breakage cosplay so.

B

This is one that I wanted to bring up. So yesterday we published a version with atom binding with the J snippets feature added to it, and so they've been a couple user reports this morning of some breakage that it's been causing, and so we knew that we were reinterpreting some of the interpret some of the strings being past a module and so British.

C

B

In terms of people actually using it, and so there's a couple of missing wanted to bring this up, where there's some example use cases in this thread and as a proposal I think which makes the most sense of basically adding right now we have module, equals string and adding raw underscore module equals string where, if you specify module, we attempt to interpret as J snippet or PM or whatever, but raw module just says. Whatever's thought all the way all the way through the West of Ireland.

B

Don't worry about it, so don't interpret it whatsoever, and so these existing use cases could switch back to using that and then otherwise pick up the GS Tippit's when they're able to, and so if anyone has any thoughts on that I can just bring that up here, but all that we can, if, if there are strong thoughts, we could chat about it here.

A

Sounds good to me.

A

D

A

Think that's a meeting unless anyone has anything they'd like to bring up would.

B

A

One item in.

B

C

A

Missed that one nice awesome okay! Well, let's do office hours.

C

The question I had is I'm trying to I'm using a crate, that's using the crate, that's using Lipsy and as part of it and I'm, trying to obviously compile two Azzam. How do you usually go about that?

C

How do people usually handle that kind of situation? It's my question should I go into more detail on much more of this. Do you know the matter my.

A

Question in response is which parts or which api's of Livesey is it using.

C

So the it's a decimal crate, so it's using CTR in 232, tu and 80 +. U +, 32 T.

C

A

Those types should work in wesam because don't end up at is some. You know raw rust type. Are you getting compilation issues? Yes,.

C

So if I do was a cargo tech to the target, wasn't it fails on that specifying those particular parts of Lipsy right.

B

So what's happening here is that for the wesson 32 unknown unknown target, the live secret is blank. It has zero contents entirely. So no types know if you guys know nothing in general, Lib C is saying whatever C library is matched up to this platform and just bind that so give you all the functions, all the types and all that and.

C

B

Of Lipsy hasn't had a equivalent to unknown target because there wasn't much of a C compiler target in it. This is a very interesting crater. Most of the time when you have these dependencies, it's rely on like some random C function like selector poll or F controller iocked allure, and so that obviously will never work.

B

But this is just types but I'm, not actually sure I'm. Looking at the crate right now, I'm not entirely certain why it's using Lib C types inside a crate as opposed to just like I ate 932. We knew 32, but so this is not a actual dependence on like functionality of the C, but just some types coming out of that, and so at least historically. The solution for this is don't depend on Lib C.

B

So, for example, the GZ like the fleet and Z lib style crate just has a back-end all stuff, and it has like a shim saying we're gonna have all the rest tapes here and use all that doesn't use lip Seattle, but longer term. We want to fill out.

B

Let's see, we were just kind of waiting on a C compiler to define what some of these types are like car, for example, can sometimes we signed or unsigned depending on the platform, and so we just kind of need, an agreement with C compilers about which is which and we're just so. We're we've been waiting on that, but now that clang 8 is being released which actually has support for the wesson 32 under target natively.

B

We can probably start matching that in terms of types and publish that, and so this create the decimal create, looks like it would just work if we publish a Lib C with those types- and we can probably just do that very soon, once the clang a is formally published and everything ok.

C

But ideally like, if I just because right now, I'm just checking because it's a dependency of a dependency before I get into my package, some stuff needs to get done. Ideally, if I just was impact that afterwards, I should be able to still I stood, be able to compile, even though the target does not match up with wasn't 32 I know no, no.

B

They're quite understood, civilian would.

C

I still be able to bundle it with wasn't pact and wasn't Benjen. Oh yeah,.

B

Once we get, this I mean the two fixes here: either send the P R to the decimal crate and stop using loops IDO, which case it'll lure or wait for a release of loop, see where it does have all the types and it'll also totally work with lesson bunch or wasn't packet was a bunch and you can just it won't, have any external dependencies or anything like that. Okay,.

C

And for that we're waiting on clone eight yeah.

B

There hasn't really been a sort of official we're waiting on this or official. This is what we're going to do, but I, don't think. Alright I think that when we actually once we actually have a fully released C compiler, which targets wisdom and has a like referenced and we can code against. We can just copy that and then we can actually add all the type so clang ages should be released. Gary any day now, I was told two weeks ago, but we'll see how that goes.

C

Thank you, yeah.

A

And you could also temporarily fork the dependency to replace the types if you just needed to. You know work around it. Real, quick, yeah.

C

C

Cool thanks thank.

A

You anyone else have anything good. Let's talk about.

A

All right well have a great week and see you next time.