[Webkit-unassigned] [Bug 41842] Add feature detection support for NRWT

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=41842





--- Comment #22 from Eric Seidel <eric at webkit.org>  2010-08-18 09:06:02 PST ---
(In reply to comment #21)
> This fall-back idea sounds good but on Qt and Gtk ports the DRT ignores this switch, gets going and blocks NRWT if we execute it, like if there only were a - switch.

Interesting.  I would have expected DRT to fail when passed switches it didn't recognize.  I figured most port would use get_opt which does all that standard behavior.

> If we want to do this I see two ways. Change all port's DRTs which are blocking with this switch, or make a more complex explicit DRT execution routine in a separate function.

In that case, I would just have a _runtime_feature_list() function which runs DRT --print-supported-features by default, catches any errors and returns None.  Then on ports which don't support it, but don't know how to error, we can override to return None manually.  How does that sound?

> What do you think about this?

I think making fancy fallback/decision logic is less important than simply separating the code paths.  The nm codepath will eventually be removed once all ports know how to do the --print-supported-features.

--print-supported-features (runtime feature detection) is important for us to move to, because it allows us to make a per-machine decision instead of a per-port decision.  For example, if the machine is not in PST, we can skip date-sensitive tests.  If the machine has a broken copy of some library, we can skip tests which depend on said library.  Both of these types of skips are currently hacked into the project in other ways.

Again, thank you very much for all your time on this.  Making NRWT better is *very* important as ORWT is way past needing replacement.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.