>I haven't seen the image diffs, so I don't know what you mean by sub-pixel. >There are two common cases: 1) a few pixels are different in corners and edges >because the anti-aliasing >chose slightly different values,
You are correct, I have seen the diffs as well and there are a few different black pixels in the edges. That's what I called "sub-pixel" mistakenly, to mean that it was subtle differences caused by anti-aliasing. Sorry for the miswording (and don't forget I am not a native English speaker ;-) ) >and 2) everything moved by less than a pixel, in which case the diffs usually >look like the outlines of entire characters in any fonts. That ones are really sub-pixels shifts. Anyway, I checked the xml DL, and the coordinates/extents are exactly the same, expect that the new skin has a bitmap, so there is no computation error and nothing we can really do about that, apart from updating the baselines. >but it now sounds like there are no tests for StageTextInputSkin? I was >going to recommend copying a few tests (probably not all) and setting them up >to use StageTextInputSkin in which case some of the old baselines would still >be valid. Ideally, you are correct, that should be done. I will do it when I have some spare time (or if there is any volunteer). But maybe we should consider StageTextInputSkin as "deprecated", because of its buggy behavior (scrolling, occlusion, etc.). So we leave it in the code for compatibility, but does it really need mustella tests ? On a side note, I don't like the idea of providing 3 skins for achieving the same result [ Editing Text in a mobile device], not because there are 3 uses cases, but because of technical limitations of each. I would really prefer to have ONE skin, improve/fix it until it behaves well in all situations, and put the other ones in the "deprecated" status. WDYT? Maurice -----Message d'origine----- De : Alex Harui [mailto:aha...@adobe.com] Envoyé : lundi 18 novembre 2013 18:22 À : dev@flex.apache.org Objet : Re: [dev] Build failed in Jenkins: flex-sdk_mustella-mobile #369 On 11/18/13 8:25 AM, "Maurice Amsellem" <maurice.amsel...@systar.com> wrote: >> the code paths these tests used to verify are still there >Sorry, I send the email too quick. > >In fact, the only visual differences between old and new baselines are >because of sub-pixel diff. The rendering is the same. And the >scrolling behavior (that is the only changed behavior) is not tested. >If it was, it would have failed. > >WDYT? I haven't seen the image diffs, so I don't know what you mean by sub-pixel. There are two common cases: 1) a few pixels are different in corners and edges because the anti-aliasing chose slightly different values, and 2) everything moved by less than a pixel, in which case the diffs usually look like the outlines of entire characters in any fonts. For 1), it might be sufficient to use png.xml files, for 2) you have to worry that even a sub-pixel shift will be noticed by someone who has "pixel-perfected" their UI. Also, I may not be understanding the situation fully, but it now sounds like there are no tests for StageTextInputSkin? I was going to recommend copying a few tests (probably not all) and setting them up to use StageTextInputSkin in which case some of the old baselines would still be valid. -Alex
