On 06/27/2012 05:01 PM, Marek Olšák wrote:
Hi Jordan,
please could you NOT change behavior of existing tests, so that it's
not seen as a "regression" when comparing different test results and
the test starts failing?
The preferred way to update a test is either by adding command-line
arguments enabling the new code or by forking the test. Either way,
there should be a new test item in all.tests.
Yeah. The primitive-restart test is fairly complicated already. In
my experience, the more complex the test, the harder it is to
debug/diagnose a failure.
Can you get by with a new, simpler test for xform feedback + restart?
I'm guessing you'll be examining the contents of the feedback buffer
rather than on-screen rendering to determine pass/fail. That's a
different approach than the current test so a new test seems more natural.
-Brian
_______________________________________________
Piglit mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/piglit
