Open mdittmer opened 5 years ago
//CC @LukeZielinski
@jgraham do you think there should be consensus on these types that their values and documentation should live in web-platform-tests/wpt
?
I mean the principled way to do that is probably by reference to mozlog i.e. have documentation stating that we support the statuses that are supported by mozlog, since that's the lowest-level component here. I don't think it makes sense to make a normative claim on the values, but we can certainly document the meanings of the ones that are actually used.
I see. wpt
consumes the mozlog
Python package. Yes, I think it would be good to document: in wpt
, "we support mozlog
status values docmented at "; in mozlog
, precise semantics of each status.
I can't seem to find source code/data that acts as the single source of truth for potential test and subtest status values that may appear in a test run report. (I can't find human readable documentation that includes all such values either, but that's a somewhat separate issue).
IMHO, there should be a piece of code or data file that acts as the single source of truth for what is essentially an enum with string representations of all "TestStatus" values and "SubtestStatus" values.
In addition to writing the enum down somewhere, we should, presumably, chase down all the code that gets or sets status values and have it interpret the legitimacy of a value through the single source of truth.
Depending on what form the source of truth takes (JSON, YAML, Python code, etc.) perhaps it can include metadata that serves as documentation for each value.