archive/servo
1
0
Fork 0
mirror of https://github.com/servo/servo.git synced 2026-05-01 07:14:51 +02:00
Code Issues Projects Releases Packages Wiki Activity
Page: Meeting 2014 10 27
Pages
Adding a new WebIDL binding Alternative Logo Proposals and Related Swag Asynchronous WebAssembly compilation project Austin Oxidation Autogeneration of style structs Basic SVG support project Beginner's guide to rebasing and squashing Benchmarking Benchmarks Bots Browser Engine Research Build Errors FAQ Buildbot administration Building for Android Building for Magic Leap Building for UWP Building on ARM desktop Linux Building CI Services we use CSS parse error reporting CSSOM student project Canvas rendering project Cargo upgrade service project Code rust concurrency Code Review Code of Conduct Coding standards Compiler upgrade recipes Compositor Layer Design Contributing Control Servo using WebDriver Creating and viewing WARC web archives in Servo Creating new OpenSSL Windows binary distributions Cross compiling from linux to mac Crowbot Css selector matching meeting 2013 07 19 DOM Design DOM documentation DOM missing pieces Debugging JS web compat issues Debugging and editing tools Debugging Design Developer tools student project Devtools CSS errors Devtools plans Devtools Diagnosing SpiderMonkey JIT issues Eric Atkinson visit 2013 09 10 Events and sundry Expand HTTP request response monitoring Fetch improvement project Firefox Reality release notes FirefoxReality build Firewall setup for servo master1 Focus student project Form validation student project GSoC project brainstorming Garbage collected DOM Getting started with layout GitHub Labels Github & Critic PR handling 101 Github workflow Glossary Governance Graphics toolkit integration HTML parser improvement project HTMLElement binding conversion HTTP archive support project HTTP library requirements Hawaii Rooting High priority content for layout Highfive HoloLens 2 test plan Home How to generate GStreamer binaries for CI Image load conformance student project Image maps project Implement HTML charset parsing project Implement ImageBitmap project Implement missing WebAudio automation student project Implement support for missing XMLHttpRequest APIs Implement worker modules Implementing a web standard (RGSoC) Improve specification conformance of unicode bidi library Incremental flow tree construction Infrastructure Integrate xml5ever Intern project brainstorming Intern projects JS objects, wrappers, and cross origin concerns 2013 08 07 Layout 2020 Layout Overview Layout resources Layout revamp ideas Leo meyerovich visit 2013 07 22 Linux sandboxing London Oxidation London Security Meeting 2014 10 27 Meeting 2014 12 08 Meeting 2012 02 08 Meeting 2012 02 16 Meeting 2012 07 20 Meeting 2013 04 01 Meeting 2013 04 15 Meeting 2013 04 22 Meeting 2013 04 29 Meeting 2013 05 06 Meeting 2013 05 13 Meeting 2013 05 20 Meeting 2013 06 03 Meeting 2013 06 10 Meeting 2013 06 14 Meeting 2013 06 17 Meeting 2013 06 24 Meeting 2013 07 01 Meeting 2013 07 15 Meeting 2013 07 22 Meeting 2013 07 29 Meeting 2013 08 05 Meeting 2013 08 12 Meeting 2013 08 19 Meeting 2013 09 09 Meeting 2013 09 16 Meeting 2013 09 23 Meeting 2013 09 30 Meeting 2013 10 14 Meeting 2013 10 21 Meeting 2013 10 28 Meeting 2013 11 04 Meeting 2013 11 18 Meeting 2013 11 25 Meeting 2013 12 02 Meeting 2013 12 09 Meeting 2013 12 16 Meeting 2014 01 06 Meeting 2014 01 13 Meeting 2014 01 21 Meeting 2014 01 27 Meeting 2014 02 03 Meeting 2014 02 10 Meeting 2014 02 24 Meeting 2014 03 10 Meeting 2014 03 17 Meeting 2014 03 24 Meeting 2014 03 31 Meeting 2014 04 07 Meeting 2014 04 14 Meeting 2014 04 21 Meeting 2014 04 28 Meeting 2014 05 05 Meeting 2014 05 13 Meeting 2014 05 19 Meeting 2014 06 09 Meeting 2014 06 17 Meeting 2014 06 23 Meeting 2014 06 30 Meeting 2014 07 07 Meeting 2014 07 14 Meeting 2014 07 21 Meeting 2014 07 29 Meeting 2014 08 04 Meeting 2014 08 11 Meeting 2014 08 12 Meeting 2014 08 18 Meeting 2014 08 25 Meeting 2014 09 08 Meeting 2014 09 15 Meeting 2014 09 22 Meeting 2014 09 29 Meeting 2014 10 06 Meeting 2014 10 13 Meeting 2014 10 20 Meeting 2014 11 10 Meeting 2014 11 17 Meeting 2014 11 24 Meeting 2014 12 15 Meeting 2015 01 05 Meeting 2015 01 12 Meeting 2015 01 26 Meeting 2015 02 09 Meeting 2015 02 23 Meeting 2015 03 02 Meeting 2015 03 16 Meeting 2015 03 30 Meeting 2015 04 06 Meeting 2015 04 13 Meeting 2015 04 27 Meeting 2015 05 04 Meeting 2015 05 11 Meeting 2015 05 18 Meeting 2015 06 01 Meeting 2015 06 08 Meeting 2015 06 15 Meeting 2015 07 06 Meeting 2015 07 13 Meeting 2015 07 27 Meeting 2015 08 10 Meeting 2015 08 17 Meeting 2015 08 24 Meeting 2015 08 31 Meeting 2015 09 14 Meeting 2015 09 21 Meeting 2015 09 28 Meeting 2015 10 05 Meeting 2015 10 12 Meeting 2015 10 19 Meeting 2015 10 26 Meeting 2015 11 02 Meeting 2015 11 09 Meeting 2015 11 16 Meeting 2015 11 30 Meeting 2016 01 04 Meeting 2016 01 11 Meeting 2016 01 25 Meeting 2016 02 01 Meeting 2016 02 08 Meeting 2016 02 22 Meeting 2016 03 07 Meeting 2016 03 21 Meeting Devtools Servo 2 Meetings Microdata project Minutes Hackathon 2012 03 27 Missing DOM features project More ServiceWorker support project More developer tools student project Mozlandia Automation Mozlandia B2S Mozlandia JS Mozlandia Rust In Gecko Mozlandia WPT Mozlandia gfx Mozlando Devtools Servo Mozlando Oxidation Mozlando SM Servo Mozlando Servo Bluetooth Mozlando Servo MagicDOM Mozlando Servo SMStrings Mutation observer project Mutation testing project NCSU student projects Network security project Off main thread HTML parsing project Offscreen canvas improvements project Offscreen canvas project Orlando Oxidation 2018 Oxidation 2015 11 05 Persistent sessions student project Preparing ARM libraries for CI Priority of CSS properties Priority of DOM implementation Priority of dom bindings Private browsing student project Profiling Project proposal deadlines Prototype JS form controls student project Prototype ways of splitting the script crate Publishing a new ANGLE NuGet version Publishing a new app store release Push vs Pull for caching Random web content project Refactor GLES2 student project Refactor bluetooth support student project Remaining work Removing push notifications from IRC hooks Replace C libraries student project Report new contributors project Representation of computed style Research Reviewer Roadmap Running Web Platform Tests on Servo Rust HTML parser Rust SpiderMonkey debugger API Rust cssparser code walk 2013 08 02 SaltStack Administration San Francisco Oxidation Servo Benchmarking Report (December 2024) Servo Benchmarking Report (November 2024) Servo Benchmarking Report (October 2024) Servo Layout Engines Report Servo and SpiderMonkey Report Servo for Gecko Developers Specification Links SpiderMonkey related tasks SpiderMonkey infodump SpiderMonkey upgrade details Storage student project Streaming webassembly student project Strings Student project brainstorm Student projects Styling overview Stylo hacking guide Summer of Code 2014: Implement XMLHttpRequest Summer of Code 2016: Fetch API Summer of Code 2016: File support Summer of Code 2016: ServiceWorker infrastructure Summer of Code projects Summit meeting 2013 09 09 Support WebDriver based tests project Syncing web platform tests (WPT) TaskCluster Testing Tools Tracking intermittent failures over time project Transcription Notes from Servo Architecture talk in Suwon Transcription notes from rust patterns talk in suwon Transcription parallelism Transcription rust concurrency Transcription rust runtime Transription layout and acid2 Trinity College Dublin student projects UPenn student projects Updating the Rust compiler used by Servo Upgrading non taskcluster linux CI machines Upgrading the UWP gstreamer binaries Upgrading the windows LLVM binaries Upgrading wptrunner Using DOM types Using Rust Spidermonkey Prototype Using WebWorker Prototype Version 0.1 Videos and presentations WebAudio JS interfaces student project WebAudio nodes student project WebCompatBug WebSocket student project Webdriver student project Webdriver tests student project Webrender Overview Whistler 2019 notes Whistler Bugzilla Whistler FFOS Whistler GFX Whistler Houdini1 Whistler Houdini2 Whistler Necko Whistler Oxidation 2019 Work items for new contributors Workweek COW DOM Workweek alt js Workweek android arm Workweek boot 2 servo Workweek compiler lints Workweek displaylist Workweek dogfooding Workweek encoding Workweek generated content Workweek governance Workweek graphics stack Workweek graphics toolkit Workweek incremental layout Workweek js bindings status Workweek layers Workweek layers2 Workweek pixels Workweek rasterization Workweek reftests Workweek roadmap Workweek script crate Workweek security Workweek string interning Workweek tables Workweek writing modes XML parser student project infra triage notes jQuery status webxr.today support
1 Meeting 2014 10 27
Manishearth edited this page 2014-10-27 17:57:46 -07:00
Table of Contents

Servo Meeting 2014-10-27

Reminder

Please update your status at: http://benjamin.smedbergs.us/weekly-updates.fcgi/

Agenda

  • Final pass over agenda items / planning for workweek (larsberg)
  • https://etherpad.mozilla.org/Servo-workweek-2014-11
  • acid2 (jack)
  • Intrinsic sizing of orthogonal flows (simon)
  • string-cache and html5ever performance update (kmc)
  • glutin update (gw)
  • Quick infra update on split mac builds (larsberg)
  • Parallel performance benchmarks? (larsberg)

Attending

  • larsberg, pcwalton, Manishearth, jack, gw, brson, mbrubeck, samsung, simonsapin, cgaebel, kmc

Work week

  • larsberg: The only thing I wanted to mention was that we should take one more pass over the agenda, since we need a final schedule for what we're going to talk about when. So if you have anything that's not on the agenda yet, get it in soon.
  • jack: Is mrobinson going to be there the whole week?
  • larsberg: Yes, during business hours at least.
  • kmc: ride-sharing for the workweek?
  • jack: Add a section to the etherpad and we can have people sign up if interested. There are several from the Samsung OSG coming from the valley as well.
  • larsberg: Did we get arrangements on having a bus take us to the SF office for the day?
  • jack: I requested it, but I'll ping her again. The Caltrain is also something we could take on Thursday to and from the SF office, including one back to the south bay that is OK after the meetup ends. Please do add anything you want to discuss, as we will put the schedule together this week.

ACID2

  • jack: People in the community have pointed out that ACID2 has regressed. What is the status here?
  • gw: I looked at it a couple week ago. There's only one issue remaining - that the nose is offset by one pixel. I looked into that, and there's an accuracy issue with rendering borders. I haven't looked into it. It is pixel perfect on Mac and Linux otherwise, and the reftest is running on both now.
  • jack: On master?
  • gw: Yes.
  • pcwalton: I think some of the issues may be related to graphics drivers. Sometimes no other layers will show up with some devices or races with the compositor. I don't know of any races in the compositor right now, though. We could chalk it up to graphics driver issues, though I do not want to use that as an excuse. That said, if community members are seeing them, we should have them take screenshots and file them. Probably two issues: the one-pixel accuracy and one on layers.
  • mbrubeck: One is just noting that the ACID2 issue is grossly out of date.
  • gw: I should clarify: The ACID2 bugs that I fixed were related to the noscroll reftest that SimonSapin added, which just has the face and not the link above. I think when you have the full test, you might have some issues related to layers that mrobinson was going to look at.
  • jack: Let's get those bugs filed.
  • gw: I think mrobinson may have already fixed those.
  • jack: Sounds like some minor bugs to file plus the tracker.

Intrinsic sizing

  • simonsapin: I will send e-mail, since it sounds like a transmission across the vastness of space...

String-cache and h5ever performance update

  • kmc: I did some event logging so that we can look at how we're interning atoms. We are handling 80%+ with no locks at all, so the various tricks help. On google,com, it's 95% because they minify things. That is also an interesting piece of performance advice to give to web developers. I also measured the overall time of the parse_html function before/after switching. It got a little bit slower, though not much on release builds. Almost all of the time spent during parsing is DOM manipulation, as parsing itself takes no time at all. Something about our DOM bindings being slow or perhaps we are doing reflows during parsing.
  • jack: Debug mode is probably a good catch, since I think we were always compiling netsurf with release mode. It would be good to maybe start a wiki page on "how to make your pages fast in Servo." This observation about minification would be good.
  • kmc: Yes, we also have lots of advice on parallel layout. Ideally, it would be part of the Servo DevTools, and they would give feedback e.g., that we were unable to intern strings. Noticing when we can't clear floats, etc. seems like a great thing for our tools.
  • jack: I think we can probably generate those tools features from this list of best practices.

glutin

  • gw: I've been working on integrating glutin, a pure rust alternative to glfw, which would let us do headless rendering without any X dependency. It should improve the reliability of our CI infrastructure. There's one race condition on Linux where 1 in 15 runs of Servo don't render correctly. If I can get it done on OSX and headless, it should be set to go in.
  • gw: The viewport renders over top of the window title bar, and you lose the height of the title bar at the bottom. It's really strange. I've traced out the exact GL/GLX commands and it's no different when we work or not. I suspect it's got to do with ordering of those function calls.
  • kmc: I used xtrace on Servo, but never found it helpful for crashes in Skia.
  • gw: I'll take a look.
  • kmc: I think the headless testing is just the start of cool things we can do! If I understand correctly, we'll have a mode that just renders into a buffer using software GL. That makes our portability much better, we can work around graphics driver bugs, and it's just a great fallback.
  • gw: You literally get back a Vec. Only dependencies are on Mesa.
  • pcwalton: Have you integrated that with the inversion of control in https://github.com/servo/servo/pull/3761
  • gw: No, I thought that had already landed.
  • pcwalton: Main thing it uses is the ability to post an event to the OS event loop from another thread. Currently, we poll. That might be part of the cause of some of the timing issues you're seeing. So, I have a branch that does that, but it requires that functionality and I don't know if it's available in glutin yet. You can search the branch for it.
  • gw: It's not there, but I've made about a dozen PRs to glutin, and they're quite open to taking our patches. I did r+ that PR of yours, but it had a compile error.
  • pcwalton: I'll look at it later. If you want to see the code for doing that on various OSes, you can look at what I cherry-picked from glfw upstream. It is fairly straightforward.

Split builds

  • larsberg: We've split builds on Mac now. We also run wpt in a single process; which gets rid of most of the intermittents. Still some unsolved problems with intermittent failures, including different sets of failures on Mac than on Linux.
  • jack: We need to get the right people together at the work week to find solutions. If the Python web server is buggy, can we just use a better server? Most of these failures are in tests that are just serving static files, no server-side logic.
  • manish: Most of XHR is static files, but some of it looks at some specific parameters for special responses.
  • jack: I can help track that down; it's probably more important than newbors.

Parallel performance benchmarks

  • larsberg: It sounds like CSSWG is not quite convinced that parallel performance is important and viable. pcwalton has been working on DOM mutation benchmarks. Are there any other things we should work on to help make a compelling case for parallel layout?
  • pcwalton: dherman probably has some opinions here. I know he didn't want us to make too much noise about benchmarks at this point. We want to be conservative with what we claim, and also I don't want to really start an arms race at this point. I think we need to wait and hear from SimonSapin what the mood at CSSWG really was. It sounded like some people might be receptive to our arguments. Also we need to look closely at the changes they're proposing; I don't know for sure how they will affect our parallel layout.
  • IRC log: http://log.csswg.org/irc.w3.org/css/2014-10-27/#e484973
  • jack: What information does CSSWG need from us?
  • larsberg: I don't think they asked for specific info; it just sounded like people in the working group weren't sold on parallizability as a requirement in general.
  • jack: While Dave says we shouldn't be publishing benchmarks, I think he means to a general audience. I think it would be appropriate to share data with a specific audience like the CSSWG. We need to convince them that parallel is important, or it will make our job harder. I'll add this to the work week agenda. And Simon will be there in person.

Timezones

  • mbrubeck: For people in the US and Canada, this meeting (the "late" one) will be one hour earlier starting next month (4pm PST instead of 5pm PDT).
  • jack: Europeans, too. Though still the middle of the night.