02:49
<Domenic>
I tested on all the specs, but I still don't track build messages in the testsuite, that'll be fixed this quarter. So Fetch builds an identical output, it just has a different set of messages. (Apologies for that.)
I'm a bit confused by this, with a FATAL ERROR there is no output, so shouldn't the difference between the empty string and the actual fetch spec be detectable?
03:03
<TabAtkins>
No, I build the test suite with -f. Otherwise many, many specs would have no output.
04:25
<Domenic>
Could you build WHATWG specs like they are built by the WHATWG, instead?
21:42
<TabAtkins>
No, because that would mean if something became a fatal error I'd have zero output. What I will do, as I just said, is capture build messages in the test artifacts, so I can tell what else is happening.
21:43
<TabAtkins>
(For example, the testsuite does not capture or use your make script, so the sha macro y'all use in your boilerplate doesn't match. That's a fatal error! But it's also irrelevant for testing purposes, and I have plenty of macro tests elsewhere.)