Hi,
The reason for the change is that the expectations file / baseline directory matching rule is complicated and sometimes problematic when --additional-driver-flag specifies multiple flags. --flag-specific can avoid the problems. Some new tools, e.g.
remove_stale_expectations.py, support --flag-specific only.
Note virtual suites are more suitable for testing small sets of web tests with flags. See
this doc for more details. Also note that for a blink runtime enabled feature that just adds an API without skipping any existing code path, setting status: "test" or status: "experimental" in
runtime_enabled_features.json5 is preferred to a virtual suite or a flag-specific configuration.
Xianzhu