-
Notifications
You must be signed in to change notification settings - Fork 4
/
Copy path2023-11-07.md
99 lines (78 loc) · 5.32 KB
/
2023-11-07.md
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
# wpt RFCs/infra sync
* Notes archive: https://github.com/web-platform-tests/wpt-notes/tree/master/minutes
* RFCs: https://github.com/web-platform-tests/rfcs
* Infra issues: https://github.com/web-platform-tests/wpt/issues?q=is%3Aopen+label%3Ainfra+-label%3Adependencies
* Matrix: https://app.element.io/#/room/#wpt:matrix.org
## Agenda Items for Nov 7th, 2023
### RFC #165 expected-fail-message in test metadata
* https://github.com/web-platform-tests/rfcs/pull/165
* Any objections to merging this?
* Implementation? https://github.com/web-platform-tests/wpt/pull/42406
* Merged.
### Add an RFC for add_top_level_completion_handler
* https://github.com/web-platform-tests/rfcs/pull/168
* Per process this could be merged. (Needs more discussion)
* Not sure why these aren't detected as tests?
* Could be if the files are in support subdirectories?
* Original test loads a resource file.
* Many cross-origin tests do this.
* Can we not run the callbacks if we're in a non-test directory
* That will break RemoteContext
* Only need this when you can't tell where the message comes from. In WebDriver case we only listen in the top BC
* Could also just modify existing callbacks?
* window.opener can be replaced.
* Other option would be to add an argument to the completion handler callback is_top_level, rather than adding a seperate API.
* [jgraham] I think I mildly prefer that.
* Do we set this based on RemoteContext or window.top / window.opener.
* Might need to check what breaks tests.
* We have tests that check you can delete window.opener.
* Can we do the window.opener check before tests run?
* Yes. WebKit testrunner is doing this.
* Sounds like we should make an argument for the completion callback, and set it according to what works in WebKit.
* Might just be able to check for RemoteContext, but will need to ensure tests aren't doing something strange.
### RFC 169: WebRTC Echo server
* https://github.com/web-platform-tests/rfcs/pull/169
* Might need some WebRTC specific input
* No obvious red flags
* Binary dependency?
* Those are a problem because they can't be vendored.
* C dependencies are a concern e.g. for brining up new platforms
* Precednet is to make all binary wheels opt-in features with command line flags
* For aioquic there has been a problem where it didn't support new Python versions.
* Also a problem for standing up new platforms.
* To what degree does this actually need to care about codecs for the tests we want to write? Could we take a simpler dependency that only covered testing use cases (especially if it's Python-only).
* aiortc allows actually processing the video / audio data, which we perhaps don't need.
* Is this a different argument for aioquic?
* Crptography is required, but that's already used for lots of the Python ecossytem, so may be a sunk cost.
* With aioquic, in retrospect we would have preferred pure Python if possible since we ran into problems with it not being updated.
* gsnedders & jgraham to comment on RFC
### Remove tests that rely on undefined testing APIs
* https://github.com/web-platform-tests/rfcs/issues/172
* Ms2ger: from the title it sounds like this makes sense. Do we need to review exactly which tests need to be removed?
* gsnedders: RFC is probably required so we aren't deleting tests that people care about.
* jgraham: Agreed. These tests can't be in the repo if they can't be tested with standardised APIs, and it can cost other vendors money to (not) run them. But we should have an RFC with a specific list of tests so that people who are running them have a chance to move them outside the wpt repo for their local CI.
* gsnedders: For new additions I'd like to be able to revert so we don't keep adding these things.
* jgraham: Maybe we should have a lint and require codeowners approval for new items?
### Proposal: support multiple <meta name="fuzzy">
* https://github.com/web-platform-tests/rfcs/issues/164
* wpt.fyi needs to allow different amounts of fuzziness per-platform
* status quo is that this lives in the test.
* Could use in-tree metadata that applies to wpt.fyi runs
* gsnedders: that would require duplicating metadata on import.
* jgraham: Could also be that vendor CI has different fuzziness and so we don't want to import it by default.
* gsnedders: We try to make generic annotations that work for different OS versions
* jgraham: Currently you have to write fuzziness that applies to all OSes and browsers, which is even more generic
* gsnedders: That is what we're currently doing, and it affects Interop scoring.
* jgraham: Anyway, no real objections to this proposal, but it does require someone to do the work to implement (and probably requires an RFC)
### Avoiding timeouts in crash/reftests where a feature is unimplemented
* https://github.com/web-platform-tests/rfcs/issues/162
* For crashtests might need to manaully update tests to check if the test is trying to use an undefined API
* For reftests we can try making an unhandled exception set the test status to error and see how much breaks.
* Houdini tests are an example of something that are going to throw an exception when the API doesn't exist so that change would cause them to go from FAIL to ERROR.
* jgraham: I think FAIL to ERROR should be OK.
### Contact points for each browser
* https://github.com/web-platform-tests/rfcs/issues/140
* gsnedders: Please look + comment on it
### Next meetings
* Jan meeting pushed back to Jan. 9th
*