chromium: Type each character as the screen changes

[kalikiana]

See: https://progress.opensuse.org/issues/109737

[kalikiana]

openqa-clone-custom-git-refspec #15173 https://openqa.opensuse.org/tests/2443233
Created job #2449803: opensuse-15.3-DVD-Updates-x86_64-Build20220701-1-install_with_updates_gnome@64bit-2G -> https://openqa.opensuse.org/t2449803

[okurz]

https://openqa.opensuse.org/tests/2449613 took 2m26s for the chromium test module. Your verification run took 3m44s. I don't think the test should be slowed down that significantly. Please check runtime of the module.

[kalikiana]

https://openqa.opensuse.org/tests/2449613 took 2m26s for the chromium test module. Your verification run took 3m44s. I don't think the test should be slowed down that significantly. Please check runtime of the module.

Any suggestions on how to avoid that while still verifying every typed character? I didn't change anything else.

[okurz]

https://openqa.opensuse.org/tests/2449613 took 2m26s for the chromium test module. Your verification run took 3m44s. I don't think the test should be slowed down that significantly. Please check runtime of the module.

Any suggestions on how to avoid that while still verifying every typed character? I didn't change anything else.

I am not sure if my observation really shows a problem. So to be sure that your changes actually cause this slow-down can you try to run multiple tests with just the "chromium" test module included to check? E.g. call openqa-clone-custom-git-refspec https://github.com/os-autoinst/os-autoinst-distri-opensuse/pull/15173 https://openqa.opensuse.org/tests/2450411 SCHEDULE=tests/boot/boot_to_desktop,tests/x11/chromium with https://openqa.opensuse.org/tests/2450411 being "opensuse-Tumbleweed-DVD-x86_64-extra_tests_gnome@64bit" which uses one of the latest Tumbleweed pre-installed qcow files.

The option wait_screen_change seems to be a sensible approach, maybe we just need to tweak it a bit.

https://gist.github.com/okurz/295e2ad3a13a37a02294f0c6c26fe69a is the section of the log where the entering and waiting happens.

Just the first two characters:

[2022-07-04T17:59:25.030514+02:00] [debug] tests/x11/chromium.pm:33 called chromium::type_address -> tests/x11/chromium.pm:20 called testapi::enter_cmd
[2022-07-04T17:59:25.030652+02:00] [debug] <<< testapi::wait_screen_change(timeout=10, similarity_level=50)
[2022-07-04T17:59:25.032396+02:00] [debug] <<< consoles::video_base::type_string(max_interval=250, json_cmd_token="rvBsdwgL", cmd="backend_type_string", text="c")
[2022-07-04T17:59:25.070389+02:00] [debug] waiting for screen change: 0 1000000
[2022-07-04T17:59:25.577352+02:00] [debug] waiting for screen change: 0 1000000
[2022-07-04T17:59:26.080454+02:00] [debug] waiting for screen change: 1 21.236376031981
[2022-07-04T17:59:26.080613+02:00] [debug] >>> testapi::wait_screen_change: screen change seen at 1
[2022-07-04T17:59:26.080818+02:00] [debug] tests/x11/chromium.pm:33 called chromium::type_address -> tests/x11/chromium.pm:20 called testapi::enter_cmd
[2022-07-04T17:59:26.080961+02:00] [debug] <<< testapi::wait_screen_change(timeout=10, similarity_level=50)
[2022-07-04T17:59:26.081879+02:00] [debug] <<< consoles::video_base::type_string(text="h", json_cmd_token="cUPMDVMb", max_interval=250, cmd="backend_type_string")
[2022-07-04T17:59:26.118460+02:00] [debug] waiting for screen change: 0 1000000
[2022-07-04T17:59:26.621285+02:00] [debug] waiting for screen change: 0 1000000
[2022-07-04T17:59:27.125229+02:00] [debug] waiting for screen change: 1 22.3299081718199
[2022-07-04T17:59:27.125328+02:00] [debug] >>> testapi::wait_screen_change: screen change seen at 1

the "waiting for screen change" is done every 500ms and the screen change is only detected in the third iteration. I assume here we would benefit from something similar to the "no_wait" in https://github.com/os-autoinst/os-autoinst/blob/master/testapi.pm#L378 . I suggest to check the implementation of the wait_screen_change option and how that is passed to lower levels, then check the implementation of "no_wait" and then extend to make "no_wait" accessible to all users of the "wait_screen_change" option.

[kalikiana]

Waiting on the results of #2109 for now to confirm how to continue

[okurz]

Waiting on the results of #2109 for now to confirm how to continue

You don't mean pull request #2109 in this repo, what are you waiting for?

EDIT: Oh, you mean os-autoinst/os-autoinst#2109, right

[kalikiana]

Apparently looking into optimizing the underlying code is far from trivial, so I'd like to wrap this up as-is. We know it works. See also my comment on the ticket.

[okurz]

As https://progress.opensuse.org/issues/109737 still doesn't have auto-review or anything let's approve as is despite the slow-down. Created a new ticket https://progress.opensuse.org/issues/114412 to optimize the performance.