WebDriver/Meetings/2019-02-18

Agenda

[whimboo] We see again slow progress especially on updates for decisions made during last TPAC. So who would mind to add the WebDriver spec content for HTTP authentication and file uploads? I'm happy to get the code/tests implemented when time permits.

File bugs to get the start-up notifications for Marionette implemented for Firefox/Fennec and Thunderbird (whimboo)
Talk to foolip about the Selenium log API and if they would like to spearhead standardising it (AutomatedTester)
[done] Get the new window method implemented for the Python binding (whimboo)

AutomatedTester: chromedriver is planning to switch of logging because it is not in the spec.; They don’t have any data, so they are looking to see if they can get any data on it.; They will likely move this across to the WPT stuff, and that will describe it.; foolip keeps referring to CDP, but that doesn’t help us.
ato: What are they trying to do exactly?
AutomatedTester: When they switch over to W3C mode, they intend to disable it.
whimboo: Did they have concerns that they will be blamed when the logging feature gets disabled?; His general reaction was “why are people using this?”; I don’t think he fully understands that there are people within Google, for example Angular, using it for testing.; Because you can get performance logs from it.; And Facebook use it for React testing, when they’re testing with Chrome.; I need to actually have a meeting with Filip because we keep talking past eachother on a number of important aspects.
ato: I guess this doesn’t really settle the issue for us whether to implement or not to implement the log API.
AutomatedTester: No, I think the best thing we can do is to wait.

whimboo: We are not yet able to correctly initialise Marionette with the new notifications.; We only got it working with some extra workaround patches.; So once we get all this working, I will start filing bugs.; Hopefully Nick will have time this week to have this investigated.

whimboo: Yes, I’ve done this, and it will be part of the next Python binding release!; Is there anything we need to discuss about the upcoming Selenium 4.0 release?
AutomatedTester: simonstewart wants it to be a pure drop-in.; 2.5.3 is still used by users, so he has concerns that people won’t switch over unless there’s some form of backwards compatibility warning.; I’ve been moving code around a bit, but keeping the signatures the same. Then putting in lots of deprecation warnings.; I hope to rip out old cruft once we bump the version number to 4.x.; Lots of cleanup.; There’s a branch on the Selenium GitHub repo where you can see roughly where I’m going with things.

whimboo: We still have some results from TPAC that has not been documented yet, especially HTTP authentication and file upload endpoint.; Neither of these have been described in the specification yet.; I am happy to implement these features, but I would not be comfortable to do this without specification prose.; I wonder how we can unblock this so we can get a bit of traction for these commands, and get them to users.
ato: We fixed the compatibility with interaction with hidden file upload elements, so the behaviour between Firefox and Chrome should match now.; But HTTP auth is something people have asked for, for quite a long time.; I don’t have time to work on this, but I would be happy to mentor anyone who would like to look at it.; The file upload endpoint would be the easiest.; [explains how it needs to work]; I can’t remember how the HTTP auth was supposed to work?
whimboo: There was meant to be a capability, and then based on that you would enter this information into the dialogue when it appears.
ato: Yes, it’s coming back to me now. And there was a big discussion about whether it was supposed to be reactive or proactive or something.
whimboo: Yes.
AutomatedTester: I vaguely remember brain volunteered to work on this?; Could you ask brian if wants to do this?
whimboo: I will add a comment to the issue.

[fixed] For Mn jobs we use the "preflight_run_cmd_suites" steps to set a unique screen size, and to disable various Windows notifications and the screen saver (Bug 1520175) - needs to be also done for wpt jobs
[waiting for bug 1520062] Intermittent Assertion failure: m_refCount == 1, at /builds/worker/workspace/build/src/js/src/jit/ExecutableAllocator.cpp:51 (Bug 1486101)
- This is the #1 orange factor intermittent as of now, and only happens for Mn jobs
- With my minified testcase the leaks have finally been investigated
- Leaks appear due to some networking code
[needs feedback] Marionette client hangs 1000s due to channel errors (Bug 1521447)
- There is a hang of Firefox when `ProcessChild::QuickExit()` is called.
- The main process never exits and as such is causing this timeout in mozprocess
Intermittent QuantumRender TEST-UNEXPECTED-TIMEOUT | /webdriver/tests/<random>.py | expected OK (Bug 1526678)
- Slow network behavior when performing commands like navigation, but even for setting timeouts
Probably a wpt downstream sync removed a couple of skips for wdspec tests (eg. bug 1418778)

(Spoken status updates in bold.)

ato
- Investigated 1525659 which hardcodes geckodriver’s HTTPD host to the IPv4 layer
  - Found that several CLI bugs with geckodriver that has to be resolved first, so at least I laid the groundwork for fixing it (1526938)