On 2/3/20 1:06 PM, Francois Gouget wrote:
On Mon, 3 Feb 2020, Nikolay Sivov wrote:
Does any of those tests actually fail? Index access for family/font objects should never fail for valid index, other tests you changed might but on extreme conditions like allocations errors and such. I think additional trace() calls are not great, and should instead print family/face names for failed tests directly, from ok() calls. Tracing indices also becomes useless when you normally run without verbose mode.
That makes each ok() call more complex (longer variable names + wine_dbgstr_w()) but it's doable.
I would be interested in fixing failures you're seeing though, and I don't see anything new on the test page.
Not failures but tons of todos on Linux which is really the same: https://test.winehq.org/data/tests/dwrite:font.html https://test.winehq.org/data/bdf8d94e2a40f82c618ba5587cc82382c5230ac0/linux_...
Now those are mostly because of unimplemented functions so which font they are applied to does not really matter. But still, not having any indication of the parameters that caused a test to fail in a loop is wrong.
I see. I sent some patches to get rid of todo's by implementing this method.