3

Meeting 2015 02 09

Lars Bergstrom edited this page 2015-02-10 15:30:28 -06:00

Table of Contents

• Servo Meeting 2015-02-09
• Agenda items
• Attending
• Servo hiring
• Rustup
• Unused_import / unused_variable
• Rust master tracking for deps
• Cookies
• Reviews
• Windows
• Async dom APIs

Servo Meeting 2015-02-09

Agenda items

• Servo hiring (jack, larsberg)
• Yet more rustupping - https://etherpad.mozilla.org/rustup (larsberg)
• unused_import, unused_variable (jdm)
• rust master tracking for deps (manish)
• windows (manish)

Attending

• jdm, mbrubeck, larsberg, jack, savago, Manishearth, Ms2ger

Servo hiring

• jack: We have a backfill req. Please suggest awesome people to us! Low number of recruiters at Mozilla right now, despite the fact that there are 100 reqs being opened. Very little help for the first one. Lars is reaching out, but need more here. Not open yet.
• larsberg: Description - https://etherpad.mozilla.org/iiCg0nahOa
• jack: Similar to previous listings, though we had content + layout + gfx ones. This one is more DOM-y, though we're open to anybody who is really awesome. Intern stuff is also winding up. Two programs: normal (just 5 slots) with relocation to SF, MTV, or TOR offices. Probaby no Servo candidates for that. Two candidates, but under the researcher program for remote interns with remote mentors, etc. Should have more spots for the researcher program (13 total?) available.
• larsberg: Ops engineer is probably not going to get opened at least until Q2, possibly later.

Rustup

• larsberg: This seems to be going nicely. Manishearth can chat more about it. We've been plodding along, opening bugs as we run into them. I've built 7 snapshots now for this one, and we're only ten days in. We've needed lots of compiler/stdlib fixes. I think it's very close. We almost have Android building. I think unit tests pass.
• Manish: We have a couple of reftest failures, which I don't know how to fix. I filed bugs for some other errors. Android is having some problems.
• larsberg: If you can take a look, that would be great. Every time I build a new snapshot we pick up more Rust changes and need to update more dependencies. It's a bad cycle.
• https://etherpad.mozilla.org/rustup
• jack: I didn't read notes from the last Rust meeting, but I'm curious: Is the next release going to be a beta, or is there going to be another alpha?
• brson: We do think we're past the end of the library breakage; we just haven't had enough bake time to be 100% sure.
• jack: I'm looking forward to getting to the point where the syntax no longer changes. There's light at the end of the tunnel, right? The upgrades actually weren't that bad for a while, but the last couple have been tricky because of bugs introduced in the 1.0 rush.
• jack: How old is the snapshot we're porting to?
• larsberg: Yesterday afternoon.
• jack: Do we have anything queued up after that that'll need another upgrade?
• larsberg: We'll see.

Unused_import / unused_variable

• jack: I'm conflicted.
• jdm: What's your conflict?
• jack: Ending up with lots of extra lines...
• jdm: They'll show up as warnings.
• jack: Just prevents compile? Then I'm fine.
• mbrubeck: We've been good at being warning-clean. They're warnings by default, but we had opted into making them errors.
• jack: Any objections? Nope. Go for it, jdm!

Rust master tracking for deps

• jack: Manish started this.
• larsberg: He got disconnected; will be right back.
• jack: This is a good idea in general. As rust matures, we have more people pulling in Servo deps.
• Manish: need help with osx though
• jdm: Could we write mini-servo in rust-cocoa?
• jack: Probably, but would need all of UIKit. We only have the obj-c runtime stuff - sending messages to a selector, enumerating them, etc. Also rust-core-foundation with lots of libraries. But not rust-uikit. Need a whole bunch of other system libraries that don't have bindings. Need lots more in rust-core-foundation.
• larsberg: bjz has been adding them.
• mbrubeck: Would also need rust-cef.
• jack: Yeah, not too far off. Probably easier on GTK side because C bindings are easier than obj-C.
• mbrubeck: I think there are several Rust GTK bindings out there.
• jack: Could also use the "UI totally in opengl" thing?
• jdm: Conrad?
• jack: Conrod, I think. Though who knows if they have a text edit control...
• jack: Manish, did you need help with anything?
• manish: Just OSX, which I mentioned above.
• jack: I can.
• manish: BTW, in the rustup notes, I see something about a light at the end of the tunnel... probably not for servo because of all our unstable apis and literally every feature gate enabled.
• jack: Maybe not used in many of our dependencies? Because they're under pressure to be on stable?
• manish: Lots of syntax extensions, etc. Many of the other dependencies also have a ton of unstable apis in use - at least syntax extensions, lots of other things, too. They can't stabilize now, and possibly ever. (the intention is to stabilize them though)
• jack: Not that we would stop, but that hopefully the scope of rust updates would be smaller.
• manish: It's much less severe than what we used to have, at least.

Cookies

• jack: Did they land?
• manish: Yes. And they work.
• jdm: We did!

Reviews

• jack: Any other big ones? Probably still pretty tall queue...
• jdm: Better. There's sandboxing, a bunch of CSS ones, etc. Nothing particularly big, I think.
• jack: Anybody reviewing sandboxing?
• jdm: yes, slow
• larsberg: Want to make sure this works on embedded. I will r- / roll it back if it breaks Android/gonk.

Windows

• Manish: There are some community members interested on Windows. Should we start on bindings for it?
• jack: Fine with them starting on it...
• manish: What does SM need?
• jdm: I have no idea. I assume it will "just work".
• larsberg: Might be problems related to mingw requirement from Rust, but supposedly not being reliable for Gecko pieces (which require the VS stack). I'd hate to encourage people to do this work if we know there are dragons.
• jack: Big pieces are a new compositor, using ANGLES under Skia, doing event handling, etc. Maybe people could work on that? Not opposed to people working on it, but it's just not a priority. Also don't want to block patches on yet another target platform.
• mbrubeck: There is some Windows support in Glutin, at least.

Async dom APIs

• larsberg: Can ms2ger comment on the reaction to new asynchronous DOM API speculation?
• Ms2ger: I don't think we have the expertise or people power to start inventing new DOM APIs
• larsberg: We have open reqs, if anyone knows someone who can work on this.
• Ms2ger: And since we don't have Promises, we wouldn't be able to implement them anyway
• jack: A reasonable point. jdm, you said we will have to make our own promises because SM won't have them until 2016?
• jdm: Yes. Concerned that basing a promises impl on a non-upgraded SM is silly, so there's a dependency there, too.
• jack: So, first upgrade SM. If that happens before 2016, then we have to build promises on top of it?
• jdm: Yes.
• jack: Ideas on when the SM ugprade might happen?
• jdm: Possibly first half. Was hoping for this quarter, but I have many other things I am tasked with by the end of this quarter.
• jack: Maybe talk about promises in Whistler / June workweek. Hopefully SM will be ready to go by then.
• jdm: Seems plausible.