Skip to content

Meeting 2016 03 21

Jump to bottom
Josh Matthews edited this page Mar 21, 2016 · 1 revision

Servo Meeting 2016-03-21

Vidyo meeting - https://v.mozilla.com/flex.html?roomdirect.html&key=aiRAKuP7pjdJ (Servo room):

  • *** Early (Monday, 12PM US Central) meeting
  • Late (Monday, 4PM US Central) meeting

Reminder: use http://statusupdates.dev.mozaws.net/ https://github.com/servo/servo/wiki/Roadmap (to inform goals: https://public.etherpad-mozilla.org/p/Servo-Q1-2016 )

Agenda items

Attending

  • larsberg, jdm, pcwalton, jack, simonsapin, nox, ms2ger, mbrubeck, edunham, ajeffrey, till

Last week

Dropping GLES 2.0 in WebRender

  • larsberg: a few vendors are still shipping products with Mali, which uses GLES 2. Samsung in particular, on their S3/Z phones.
  • pcwalton: Also rPi 3 uses GLES 2. I think that where there is a will there is a way to make it work on GLES 2. Once people are really interested in making it work, then it should be possible to implement fallback.

What's blocking WebRender as default?

  • jack: We were going to do this on the mailing list last week; instead we put it in the notes today. Discussion should probably be on the mailing list. Strawman plan is on the list. Please respond on the list!

Automating failure/intermittent finding

  • larsberg: Grepping by hand for failures and searching for intermittents is annoying. Any objections to having our automation automatically post the list of failing tests to the log, and/or auto-linking to known intermittent test issues?
  • jack: Could take the file that failed for anything with the Intermitent tag and the names of the failing tests? That would at least get them linked up and would help with scoping thes earch.
  • jdm: That's what I had in mind. There's a bitrotted PR for highfive that could interpret travis output & give you something to work with. But that needs a lot of rebasing over the highfive rewrite (https://github.com/servo/highfive/pull/26).
  • simonsapin: Travis? Not buildbot?
  • edunham: That PR doesn't, but we should be able to do the same thing. The only hard part of sniffing out the part of the logs and putting it in there is finding out how much of log we care about. Occasionally, the warning at the end can be pretty huge.
  • jack: Mainly just want test failures, which are all in one place, and grep should be able to get them out. After that, it's searching the GH API.
  • jdm: Someone commented in a mentored issue about taking the output of build jobs and doing something with it. https://github.com/servo/servo/issues/7481
  • nox: How does Firefox deal with intermittent identification and failures? TreeHerder?
  • jdm: Manual sherrifs who tag it. Goal for many years has been to automate it.
  • larsberg: They have formal rotations in Gecko, we just have folks who are interested.
  • jack: Next steps here are to get the link for the PR that needs rebasing into the notes and then opening a bug to see if any of our python contributors would like to hack on this.
  • larsberg: Even just having the names of the failing tests would be sufficient for me to be very happy (without the searching).
  • nox: Which output will it grep?
  • jack: test-wpt.
  • nox: Should we then, in buildbot, just generate a smaller log?
  • jdm: There are the raw logs (hundreds of megs) and the stdout links, which are much smaller.
  • Ms2ger: Can we not run grep on the formatted output of structured logs, please
  • Ms2ger: Relevant: https://public.etherpad-mozilla.org/p/sheriff-autostar-feedback
  • jdm: Could possibly have the buildbot do structured reduced output.
  • jack: I think we could just do it on the buildbot server, then.
  • nox: Could also run the reftest analyzer!
  • Ms2ger: We should be able to just upload another log file
  • simonsapin: Smaller note with only the failures would be really good.

Swag

  • larsberg: We have some budget! Please send suggestions for things you would like to have a Servo logo on (and possibly a newly-created one!) at London to me.
Clone this wiki locally