Meeting 2014 03 31

Servo Meeting 2014-03-31

Agenda

• demoing ACID2
• q2 agenda
• hiring / intern update
• work week update
• Android status (larsberg)
• acid2 post-mortem
  • z-index
  • tables

Administrivia

• jack: We have the the ServoSamsung Vidyo room for the 9AM KST time meeting

ACID2 demo notes

• jack: We have some notes for your demo to make sure you have everything you need to demonstrate the ACID2 materials:
• currently available at https://github.com/pcwalton/servo/tree/acid2-2 (merged to master soon hopefully)
• Does not work with parallel layout. Run as ./servo -y 1 ../src/test/html/acid2.html
• jack: Do you have it working there? Is there anything we can do to help with the demo? Have you been able to pass ACID2 there?
• ryan: Yes, we downloaded it and it passes OK
• jack: Do you need anything from us for demos?
• ryan: We will do the demo soon. When we tested it, everything was OK, so it should be fine.
• jack: Great! Let us know how it goes.

Q2 agenda

• jack: Lars and I talked about the kinds of projects our new hires and the interns might do. We missed some stuff in Q1 (incremental layout, HIDPI, nightly builds, android buildbot). In Q2, we decided to move fragmentation up into Q2 from Q3. And I believe that jdm wants to do the spidermonkey upgrade in Q2. The plan we came up with is that one of our new hires will work on cross-language inlining to get the spidermonkey upgrade going. And I believe the other one will start working on copy-on-write DOM (though probably not their first project).
• mbrubeck: I chatted with Lars a big about some that I'm interested in, but I'm open to anything that you throw at me.
• jack: lars and I had a bunch of good ideas for starter projects because we will have a lot of them
• pcwalton: I have some project ideas as well for the new hires and interns as well.
• jack: We should create an etherpad. Laleh (intern) starts June 30th. The others start on April 28th and somewhere around then. So let's collect those on an etherpad (or file issues for them). Speaking of, are we doing the GSoC stuff, jdm?
• jdm: Yes, GSoC will do XMLHttpREquest. Also, if pcwalton signs up as a mentor, we might be able to get the slot for the font work.
• pcwalton: Tell me what to do there.

Interns

• jack: Who is mentoring whom? ANybody specifically want to mentor somebody? pcwalton and jdm have one each from GSoC. (google summer of code). If anyone wants to mentor somebody, please ping me offline and we will sign you up with one of the four interns (laleh, brendan, ms2ger, and intern-to-be-named)
• simonsapin: Are these the four for research?
• jack: With GSoc, we will have 6 interns.
• jdm: I should be able to mentor another one.
• jack: Maybe ms2ger shoudl mentor us :-)
• mbrubeck: I'll mentor an intern if someone mentors me...
• simonsapin: Did they apply for positions?
• jack: They're all for servo, but they don't have specific projects yet. Generally, we start them with a small project in their area of interest and then get a bigger project. One example, last year Eric did text underlining (simple CSS stuff) and then worked on all the initial CSS float support. So a small start project and then work with the person on what their big project will be. THe one exception is Laleh, who wanted to look into power usage on mobile devices. Any other questions on interns or hiring? On new hires, we have SimonSapin starting officially full-time today. Other two hires are planned for April 21st and April 28th right now.

Workweek update

• jack: Poll results were in favor of the week of June 2nd. I will send an e-mail, but plan on June 2 for the next Servo workweek. Location still to be determined.

android status

• lars: i've gotten passed all the linker errors and pulled in the various rust hotfixes. the only thing left right now is that we make it most of hte way through layout but fontconfig can no longer find it's android specific config file. i just have to fix that, and other than that everything seems to work. i hope to have it landed in the next day or two.
• jack: how big is hte PR?
• lars: touched almost every submodule and changed stuff in every part of servo. i'm planning to wait until after acid2 and next rust upgrade land.

next steps

• jack: After the ACID2, we'll land a Rust upgrade. Then, lars wants to run a sed script to remove all of the trailing whitespace so that we don't keep seeing them in Critic.
• mbrubeck: What's wrong with it?
• jack: When we do critic reviews, people see the whitespace issues, but they're often from existing code. So, we'd like to just remove all of these whitespace issues in one go.
• simonsapin: I wish we had rustfmt
• mbrubeck: Yeah, it would be great to have a tool for this.

ACID2 post-mortem

• jack: ACID2 tested a specific set of edge cases in the spec. I just wanted to check how much of the general cases we implemented while getting those edge cases together. I'd like to prioritize or list parts of work that we didn't do getting ACID2 together. Examples are z-index and tables. I don't know what exactly is left over in the features we partially implemented.
• pcwalton: I'm worried about doing more before we have vertical writing direction, bidi, and pagination. We've been cautioned that we should not build more features until those are there. I'd like a moratorium for layout features that don't simplify, clean up code, etc. unless we know it does not affect writing modes or pagination.
• simonsapin: I plan to do writing mode next.
• jack: So you think we shouldn't implement tables and z-index until after writing mode?
• pcwalton: z-index is fine because it does not affect writing mode or pagination. I'm not sure how tables impact pagination; we should figure that out before we do much more work on them.
• simonsapin: for writing modes, mostly about using logical directions. So, if we have less layout code to convert, then later when we write new code if we use the new APIs, that easier than having to convert more of it.
• pcwalton: I agree. We should not add more until we have writing modes.
• jack: I agree in principle. Just wondering if there aren't leftover ACID2 tasks. Like stuff that would have been needed if ACID2 wasn't dumb.
• pcwalton: There's some CSS2.1 stuff, but we'd have to take it case-by-case.
• jack: Like, there's no colspan or rowspan in the ACID, but it seems wrong to say we're done with things...
• pcwalton: I just don't want to rewrite it. There's a lot of stuff we can be doing. z-index. layerization.
• simonsapin: Everything that is not actually layout (styles, graphics, displaylist construction)
• pcwalton: Probably flow construction as well. reflow is something we can probably do without impacting writing direction too much. In fact, I think doing that soon is nice because our overflow code is untested and the longer that waits the more concern I have...
• simonsapin: Overflow?
• pcwalton: During assing-heights, we determine the box that contains all of the ink that a flow contains. Need that for incremental reflow to know what tiles we have to invalidate. We try to compute it, but it's probably incorrect.
• simonsapin: I think that's affected by writing modes...
• pcwalton: Oh, okay...
• simonsapin: That said, I don't want to block everything.
• pcwalton: We can do things like fixing layers, which needs a total rewrite at this point.
• jack: One of our new hires might look at that. We also talked with a contractor this morning who has done this in webkit.
• pcwalton: is bidi independent of vertical writing? Or writing direction?
• simonsapin: RTL you need and is impacted by bidi, which allows you to change mid-line. Eg. tables are affected by 'direction' but not bidi.
• pcwalton: Can do the incremental flow construction part of reflow. And it should not mess up writing direction. Can also have more than one person working on writing direction.
• larsberg: We have plenty of Q2 work even staying out of the way of writing direction
• pcwalton: Also pagination, which we have to discuss what that entails and what we need to do to future-proof for it. I know you implemented some of it in WeasyPrint, Simon, but I don't know what pagination affects. What parts of the code need to change for that?
• simonsapin: Making sure that we support multiple fragments is key, but I don't know more specifically than that. (WeasyPrint is purely sequential without dynamic updates, takes lots of shortcuts.)

android

• aydinkim: Are you doing the android port?
• larsberg: yes
• aydinkim: I've been doing other work until now. I can help you with it. We can talk on IRC.
• lars: i'll open up a PR with everything else we need and if you could help out with fontconfig stuff, that would really help.

Next plans

• jack: What is everyone planning to work on next at Samsung?
• aydinkim: Preparing for demonstrations for the higher-level supervisors
• ryan: First the demo, then we need to talk with the rest of the managers, but it takes time to get a response. Young-Il will let you know. We are waiting for a response.