| | 1 | = Improving Interop with WPT |
| | 2 | ''by Philip Jägenstedt'' |
| | 3 | |
| | 4 | - Why does interop matter? |
| | 5 | - Users, developers, browser engines |
| | 6 | - Users |
| | 7 | - Go to website on favorite browser, but broken |
| | 8 | - Go find a browser it’s meant to work on |
| | 9 | - All the things we do for interop make our lives easier |
| | 10 | - Developers |
| | 11 | - MDN ran largest survey ever |
| | 12 | - Top pain points cross-browser compat/interop/testing |
| | 13 | - What to do? |
| | 14 | - A shared test suite, for all the web |
| | 15 | - 20 commits from WebKit |
| | 16 | - Believe it is infrastructure |
| | 17 | - High Level View |
| | 18 | - Chrome + Firefox + Safari |
| | 19 | - 0/3 ~ 5K tests |
| | 20 | - 1/3 ~ 6K tests |
| | 21 | - 2/3 ~ 6K tests |
| | 22 | - 3/3 ~ 20K tests |
| | 23 | - Moving to target the browser specific failures (2/3) |
| | 24 | - Interop is steady |
| | 25 | - Browser specific failures in WPT |
| | 26 | - Gap is growing |
| | 27 | - Methodology for data for this presentation |
| | 28 | - WPT splits into tests and sub-tests |
| | 29 | - Sub-tests are within the test file |
| | 30 | - Some tests with too many subtests |
| | 31 | - Encoding tests have absurd |
| | 32 | - Normalized to test level |
| | 33 | - Runs on Azure pipelines |
| | 34 | - STP runs every 3 hours, takes < 2 hours |
| | 35 | - Edge Dev and Canary too |
| | 36 | - Thanks Microsoft and @mustjab |
| | 37 | - Runs on Taskcluster |
| | 38 | - Chrome + Firefox on every commit |
| | 39 | - WebKitGTK ally |
| | 40 | - Thanks Mozilla and @jgraham |
| | 41 | - Room for Improvement |
| | 42 | - STP updates |
| | 43 | - Reliability |
| | 44 | - Thanks Apple and @burg |
| | 45 | - https://github.com/web-platform-tests/wpt/projects/4 |
| | 46 | - Epiphany / WebKit |
| | 47 | - Also run subtests on Linux for a second opinion |
| | 48 | - WebKit on Linux |
| | 49 | - Google partnering with Igalia |
| | 50 | - WebKitGTK stable+nightly runs |
| | 51 | - Fixing noise in the data |
| | 52 | - 2-way Sync |
| | 53 | - Script for WebKit exists |
| | 54 | - 2-way Sync in Chromium |
| | 55 | - Update impl+tests in single review |
| | 56 | - Use of WPT by Chromium devs up ~220% (*j.mp/wpt-yes) |
| | 57 | - ~55% of impl+test changes now use WPT |
| | 58 | - Chromium & Gecko |
| | 59 | - Automatic import |
| | 60 | - Automatic export AFTER changes land |
| | 61 | - WebKit |
| | 62 | - Manual import (?) |
| | 63 | - Yes it is |
| | 64 | - Manual (?) export BEFORE changes land |
| | 65 | - Room for improvement / collaboration? |
| | 66 | - Issue on export PRs? |
| | 67 | - A Wishlist |
| | 68 | - Collaborate on wit.fyi |
| | 69 | - Wish there was a way to browser-specific failures in the same way computed for the graph |
| | 70 | - Running web-driver to internal builds |
| | 71 | - Collaborate on running WPT on iOS |
| | 72 | - Collaborate on MDN survey: developers say interop is a big thing |
| | 73 | |
| | 74 | - WPT Right approach |
| | 75 | - Importing all the tests could slow us down |
| | 76 | - Maybe keep them separate? |
| | 77 | - New tests added don’t look like regressions |
| | 78 | - WPT: Tracking WPT as a sub-module is not compatible |
