Skip to content

Meeting 2015 01 05

Jump to bottom
Lars Bergstrom edited this page Jan 6, 2015 · 3 revisions

Servo Meeting 2015-01-05

Agenda items

Attending

  • kmc, samsung, pcwalton, zwarich, jdm, brson, jack, gw, larsberg

Status on CSS feature support

Accessibility

  • zwarich: Accessibility plans?
  • jdm: Similar to electrolysis was the plan.
  • larsberg: I talked to the accessibility team members in portland. They said they'd recently done it for FFOS, etc. and had some ideas about what we needed to do, though they're pretty tired from having just rushed to get it all working there.

Cookies review

  • jdm: There is a PR that has been tested on various sites that allows you to log in. Who will review it?
  • jack: We can upvote stuff on reddit?
  • jdm: Theoretically...
  • larsberg: Are there good specs in this space, or do we need to steal somebody from Gecko? Scary / security minefield to me :-)
  • jdm: Good spec by adam barth in 2012.
  • zwarich: The cookie situation was bad in 2008/2009, but now actual practice has been standardized. You can leak information if implemented incorrectly, but that's harder to do in Servo given its architecture.
  • jack: JSAPI?
  • jdm: Just the networking stuff. DOM should be easy.
  • jack: This doesn't look too bad; most is the toplevel domains.
  • kmc: Full public suffix list?
  • jdm: Yes. It's just a big static constant.
  • kmc: Yeah, should talk with gecko people about when to update that.
  • pcwalton: I had to do something similar for CSS counters. There are some traditional ways for ordering the hiragana based on a poem...
  • jack: I can look at cookies. Only 400 lines of code once you ignore the constants for the domains.

Rustup progress

  • jdm: I have a PR as of 10 minutes ago!
  • jack: gw got android working?
  • jdm: Yes.
  • gw: It's building now, yes. Separate from the glutin fixes. There's one more change I need to make. We can stick with the same rustc; I just need to add something to the rustup branch for the submodules.
  • jdm: I squashed everything, so we can either re-force-push the old branch or you can put it onto the squashed one.
  • gw: I'll just add it to the squashed one.
  • jack: The new android stuff is the only thing that needs review?
  • jdm: Nothing has been reviewed yet. Should we not do future upgrades?
  • larsberg: Most of closure reform has landed, but there are a few more things that are supposed to land before the alpha that niko thinks we should wait for. Don't see reason to argue against that.
  • jack: What precipitated this upgrade?
  • larsberg: Set of features we grabbed was everything except the really big closure reform stuff. We were 1.5 months behind.
  • pcwalton: Upcoming breakage - "T, ..N" -> "T; N". Also isize/usize, but that's not in yet. Alpha is coming on Friday.

Workarounds for 4545

  • gw: rust-png and rust-harfbuzz crash as soon as you link to anything that uses the gtk libs. Simon suggested we need to talk with the gecko team? Not sure who would be the right people to reach out to for that.
  • pcwalton: I'm not sure who is the desktop linux guru... jonathan kew (jfkthame) because he knows about harfbuzz? Maybe just ask in #developers.
  • jdm: ted or mike hommey (glandium), perhaps.
  • jack: Also, I meant to chat with acrichto about this at the workweek, but maybe this is something we need to handle in Cargo. I assume there's something firefox does that makes this not a problem.
  • gw: My local workaround is a branch of those projects and in those rust ones, I change to link to the system ones instead of the local ones. It works fine, but it seems particularly flaky since it's a hardcoded FFI binding without looking at what's on the system...
  • pcwalton: Don't you typically add a prefix (like MOZ_) to a whole bunch of symbols, and use those instead? IIRC, we do that in sqllite, because of this exact issue.
  • jack: At the source level?
  • pcwalton: At least jemalloc, and probably sqllite, probably have a configure switch and preprocessor magic for this. There may be a better solution via link flags (telling it to prefix stuff), but I don't know if such magic exists.
  • jdm: glandium is deeply involved with this stuff.
  • jack: We have the same trouble as gecko, so they should know the solution we need to use here.
  • pcwalton: I bet it's some magic linker flags.
  • jack: Last change required for miniservo-gtk, right?
  • gw: Yes. There's also a minor PR to change the resources path, and then it'll just work.
  • jack: I wonder why we don't have this problem on OSX...
  • zwarich: ImageIO uses libpng for png decoding.
  • jack: But doesn't expose the symbols.
  • zwarich: I would imagine not; it would be a terrible bug if so.

Reviews

  • jack: Let's get rustup landed, and then get on the reviews of the open PRs. Maybe a little break before the next rustup?
  • larsberg: You'll have a short break because it takes time for making the snapshot!

innerhtml

  • pcwalton: What's the status of innerhtml?
  • kmc: ask eddyb
  • jdm: And cparis
  • mbrubeck: eddyb has a HTML parser written in JS that works as a polyfill sufficient to load facebook...
  • kmc: He did a html5ever patch, too. Not spec-complaint yet. (https://github.com/servo/servo/issues/849#issuecomment-67271740 )
  • pcwalton: We will need to be able to inject polyfills for SVG.
  • manish: I'm working on that anyway - we need it for Gaia support
  • pcwalton: That would be great! Once Canvas lands, we can do SVG polyfill on top of it. There's already an existing CanVG polyfill that's used for places where SVG is compiled out.
Clone this wiki locally