Bug 45856 – new-run-webkit-tests: make port configurations data-driven, not hardcoded

• Component: New Bugs
• Status: RESOLVED WONTFIX

[Dirk Pranke]

new-run-webkit-tests: make port configurations data-driven, not hardcoded

[Dirk Pranke]

Created attachment 67748 [details]
Patch

[Dirk Pranke]

Created attachment 67749 [details]
Patch

[Tony Chang]

Comment on attachment 67749 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=67749&action=prettypatch

Do you plan on this configuration value to always be the same as options.configuration?  It might be a bit confusing to have this and options.configuration as different values (e.g., we use options.configuration to find the test binary).

Also, are you sure we want to replace Debug/Release with this?  Even if a port wants to override this, wouldn't we still want to be able to differentiate Debug vs Release builds (e.g., GPUDEBUG vs GPURELEASE)?  I guess I'm not sure we should deprecate |is_debug_mode|.

> WebKitTools/Scripts/webkitpy/layout_tests/layout_package/test_expectations.py:129
> +            tests, test_platform_name=test_platform_name,
> +            is_lint_mode=is_lint_mode, tests_are_present=tests_are_present,
> +            overrides=overrides, configuration=configuration)

It seems weird to name parameters that aren't named in the prototype (is_lint_mode and configuration).  Can we just pass the arguments in in the same order as defined in __init__?

[Dirk Pranke]

Comment on attachment 67749 [details]
Patch

per conversation w/ [e-mail], there's no real need for the list of configurations to be overrideable on a per-port basis, and there's already a list of build types in test_expectations.py. I will rewrite the code accordingly and upload a different patch.

[Dirk Pranke]

Marking as WONTFIX ... I will take a slightly different approach to this as part of the GPU expectations merge.