optimization tests vs. cache

[Simn]

1. Open tests/optimization and compile it through the server. Observe that there are no failures.
2. Change any of the @:js and compile it through the server again. Observe that there's a failure now, as expected.
3. Change the @:js back to what it originally was and run it through the server one more time. Observe that there's still a failure.

I think this must be some metadata insanity going on.

[kLabz]

does that happen with pre-hxb nightlies too ?

[Simn]

I don't think so, but I haven't tried.