On 12/27/13 12:46 PM, "Justin Mclean" <[email protected]> wrote:
>Hi, > >> I got 9 failures that way. >Odd I still get 5 - 3 are anitaliasing, 2 are due to incorrect results ie >new results are better. Are we running the same number of tests? Yup, same number of tests. The 3 "antialiasing" may just be different y positions by a fraction of a pixel. > > [java] ===================================================== > [java] Passes: 42 > [java] Fails: 5 > [java] ===================================================== > >> I'm wondering, is there a "key" change in this patch of six changes? >>I'm >> guessing it is this part: textWidth > prefWidth. >That logic was there before but wasn't always correct. eg Perviously one >of the tests truncates "Check the message embedded font." to "Check the >message". Using the measured height fixed the vertical position issue >that exists with the first patch. I'm confused now. The patch seemed to show that as new code. > >> It isn't clear that textHeight really needs caching. >It probably doesn't but would require more code changes, otherwise any >test where the icon is higher than the text would probably fail as the >text would be positioned higher. If you take out the last changes and go back to newHeight instead of textHeight does it change your results? -Alex
