1

Meeting 2015 04 27

Josh Matthews edited this page 2015-04-27 20:36:59 -04:00

Table of Contents

• Servo Meeting 2015-04-27
• Agenda items
• Attending
• Reviewable
• Monthly blog posts
• Issue tracking
• Infrastructure issues
• Whistler meetings

Servo Meeting 2015-04-27

Reminder: use http://benjamin.smedbergs.us/weekly-updates.fcgi/

Agenda items

• Reviewing Reviewable (larsberg)
• Monthly blog posts
  • Weren't we going to do that?
• Issue tracking
  • BMO now supports login-with-github
• infra issues (pcwalton)
• Whistler meetings (larsberg)

Attending

Manishearth, jack, jdm, kmc, Savago, pcwalton, mbrubeck, gw, larsberg

Reviewable

• jack: Hatred? Love?
• mbrubeck: Seems good. Learning curve is steep due to all the icons without labels.
• jdm: I played with a PR that was rebased, squashed, and review comments ALL at the same time and reviewable did pretty well. Some comments were marked in the rebase, which is not great. It's not perfect, but Reviewable is better than critic so far.
• manish: The revision selection is a bit fiddly and weird...
• larsberg: And really low-perf on Firefox.
• jdm: The larger the PR, the worse it performs in reviewable.
• pcwalton: Great test case for Servo! Could dogfood reviewable in Servo - it's a great incentive!
• jack: I liked how they were pretty responsive to a couple issues we filed. They didn't fix the way we suggested, but did add features within a week or so. Critic is not so much under active development. No big objects except for performance in Firefox. Does it handle force-pushes?
• jdm: Yes.
• manish: It just generates a revision. While a commit can be a revision, it tracks the change on its own side. I'm not sure how well it handles merge conflicts and rebases - we should check it.
• larsberg: We should get feedback from the people who aren't at this week's meeting too.
• jack: I'll send an email to the list, and we can put it on next week's agenda. If there are no dissenting voices, we can turn Critic off and switch to Reviewable.

Monthly blog posts

• jack: Anyone have ideas for blog posts?
• Manishearth: I might write one on linting.
• larsberg: I had a Samsung blog post come out last month and a second one is scheduled to come out next month, I think.
• jack: I had a demo blog post I was thinking of doing; maybe I'll get on that next week.
• jdm: I remember suggesting that we should write an up-to-date description of what we do for layout these days.
• pcwalton: I could work on that.
• jack: Layout hasn't changed radically, but flow tree construction and steps up to that have changed a lot.
• pcwalton: Changes in layout are mostly additional features like table layout, etc.
• jack: The most interesting thing to show off as a demo might be gw's servo-shell.
• pcwalton: Yeah, it also makes good screenshots and you can run it and everything.
• gw: I can write up a draft about that.
• Manishearth: I think when rust-in-gecko gets close we should prepare a blog post, even when it's not ready to ship, since we need more people working on it.
• jack: Is it close enough now with the rust-url patch?
• Manishearth: I think we need to get more of the build system in place. But whenever we want people to start poking at what we have.
• larsberg: I'll be scheduling a talk for Air Mozilla about that.
• kmc: Just a status updates about the Gecko work would be great, even before we have something to show off. Just to keep people in the loop.

Issue tracking

• jack: ms2ger wants us on bugzilla?
• jdm: Yes.
• manishearth: Strongly against this.
• jdm: Theoretically, it's now easier to file bugzilla bugs by logging in using github account.
• jack: Does anyone else want this?
• kmc: Private bugs for security. Also kinda scary that it's all backed up only in GH. But maybe not the right time.
• manishearth: Gaia does this, but their whole workflow is broken. If you want to make a PR, you do so, and then have a link to an attachment to the PR... coordinating across multiple interfaces is hard. We already have reviewable to reduce the number of different tools.
• jack: Hard to swallow that pill... jdm? We do a lot of things with GH, does that all exist in bugzilla?
• jdm: It was all there first.
• jack: Does it work as well?
• jdm: I have a separate website http://www.joshmatthews.net/bugsahoy/ to make the UI more github-like to see the easy / good first bugs. I like what we have better in terms of being able to attract and retain newcomers. Longer we put of switching, the more we have tied up in GH.
• manishearth: For newcomers, we could review on GH for them and mirror to bugzilla, but once experienced, migrate to bugzilla...
• jdm: I'm not optimistic of having different workflows.
• jack: Also won't scale. Fine for our small community, but in a little while, it will be hard.
• mbrubeck: Bugzilla has much better tools for e-mail notifications, both for send/vs. don't and custom headers for filtering.
• jdm: Can create e-mail rules to make sure you notice when things are mentored.
• mbrubeck: If we get to a point where we can't keep track of incoming bugs, we should move.
• jdm: People in gecko are shocked that you cannot say, "I want to see notifications only for all bugs related to the DOM"
• jack: I'm mainly concerned about the input side. Gettin gpeople to file bugs in bugzilla seems impossible for newcomers.
• kmc: We can fill the gap with GH issues with either a bot or 3rd party solution.
• jdm: Another one is have a bot that auto-migrates issues from GH to bugzilla.
• jack: True. Are there dedicated BMO engineers at Mozilla? And do they have spare cycles to do this stuff for us?
• jdm: They do maintenance and feature work.
• jack: Maybe the next step here is to bring out specific concerns to bugzilla. Maybe we can get bugzilla better? Talk with them in June at Whistler to see if they have ideas?
• mbrubeck: Bugzilla is also configurable per-product (Servo would be one). So the form does not have to look anything like the one for Gecko with release tracking, etc. Bugzilla also has two different forms - one for users with full access and one for new/guest users.
• pcwalton: Little worried about the PR -> Issue integration being very good - links work, etc. If we move them to Bugzilla, we lose that integration unless we also migrate PRs to bugzilla. That's a concern.
• manishearth: more bots!
• jack: Yeah, we rely on things like "fixes #x", etc.
• pcwalton: And hyperlinks between the PR and issue. bugzilla will not link to the sha of a commit, for example. I'm not opposed to the move, but just pointing out the limitations. Also, we lose EMOJI!
• jack: Probably highest priority for BMO team.
• manishearth: BMO doesn't even support markdown; I don't know how many features we can get in there. There was a GSoC project for markdown support, but I don't think it's close to merging right now.
• jack: Probably also worth talking about this next week for talking with the morning meeting Servo folks to see if we can get convergance. Also could investigate other tools. Reviewable kinda came out of nowhere. Hopefully one of htose will do the same for bug tracking!
• manishearth: Markdown did merge! (https://bugzilla.mozilla.org/show_bug.cgi?id=330707) Just doesn't exist on BMO yet.
• jdm: The mozilla one is customized and can run custom extensions, allowing us to request specific things for our product that we would not have to convince the rest of bugzillians everywhere to use.

Infrastructure issues

• pcwalton: Lots of tests were failing; patches bounced due to issues? Also, seems we have intermittent WPT failures, especially GlobalComposite...
• jdm: I file issues tracking each intermittent. That one is disabled.
• pcwalton: We can disable them? Cool.
• jdm: The thing to do here is make sure something's filed for it first. If specific tests are causing havoc, disable them, put that in the issue (don't close it!). Also, each time you retry, please leave a pointer to the relevant issue in that PR for tracking.
• manishearth: Homu ignores everything below the first line, so you can put other things after your retry (like the issue number pointing to the flaky test)
• jdm: The XOpenDisplay error is causing the most problems right now and is just on the WPT tests.
• gw: I was not able to reproduce this locally. I think it's the same one as when we started running our normal reftests headless, so not entirely sure while they're not being run that way. I will log into the build machines and have a look.
• jack: Any more git network failure issues?
• manishearth: One today, but didn't seem to be persistent.
• jdm: At least one other network failure...
• jack: Those are going to happen sometimes. Hopefully XOpenDisplay and then

Whistler meetings

• larsberg: What teams do we want to meet with? We should get on their schedules ASAP. You can send me email and I'll track people down and set something up.
• jack: Do we need anything from the JS team?
• larsberg: We can gloat if mwu's SM update lands. :) jdm, anything else for JS?
• jdm: Not that I'm aware of.
• jack: Did Promises get on their schedule?
• jdm: I heard Boris talking about it the other day, but I'm not sure.
• jack: It might be interesting to talk to the media team about EME sandboxing, since that's probably going to be on our partner roadmap. We should talk about our plan to re-use the Gecko implementation. Maybe also vlad and the VR team.
• mbrubeck: The necko people doing rust-url?
• Manishearth: Should we do something about Gonk? What are our plans for the future of Servo on Gonk?
• larsberg: From my conversations with fabrice, until we can support Gaia, which is a lot of web API work, it's difficult to take a next step.
• Manishearth: We can polyfill some of what we're missing.
• gw: We should talk to paul about browser.html if he's there.
• jack: And again in the morning meeting to get Ms2ger and SimonSapin's feedback.