On Oct 13, 2015, at 8:29 AM, Stephan Witt <st.w...@gmx.net> wrote: > Am 13.10.2015 um 16:43 schrieb Jean-Marc Lasgouttes <lasgout...@lyx.org>: > >> Le 13/10/2015 12:32, Jerry a écrit : >>> >>> Another difference is that the cached bitmaps (second line) do not use >>> subpixel antialiasing, the rendering method on OS X. (There used to be a >>> website explaining how Woz did this on the Apple II.) I don't know if >>> this affects the metrics. >> >> Stephan, this is a part I will not be able to deal with. > > I think it's the same effect, Zhexuan describes with "the font becomes very > thin and less pleasant to read on a retina display". This is a matter or > choosing an appropriate screen font. If the user wants a stronger one it > should be possible to switch to another font for on-screen rendering.
The reason the "font becomes very thin" with bitmap caching is due to the different rendering method that I mentioned in my previous post on this thread, specifically, the bitmaps that are cached are rendered without subpixel antialiasing. I believe that it might not be possible to efficiently save subpixel-antialiased characters on OS X because each character can be rendered in different ways, depending on its placement on the screen--OS X characters are not constrained to a grid. I don't know if the following is a well-kept secret, but on OS X, you can zoom the screen to the position of the cursor by holding Control while executing a two-finger swipe upward (and zoom back out by swiping down). This is how it becomes easy to see how text is rendered. To achieve this, System Preferences -> Accessibility -> Zoom then check Use scroll gesture with modifier keys to zoom, and uncheck Smooth images. Using this zooming, it is quite easy to see how characters are rendered, and to find instances of a character that are rendered in different ways. I believe that Microsoft finally defaulted to subpixel rendering a few years ago but they might still constrain characters to lie on a grid which if true would make caching in that case a reasonable exercise, I suppose, since each instance of a character would be the same. So, (further speculating) if the cached characters are constrained to lie on a grid while the subpixel versions are not so constrained, I would guess that that could affect the width of words. Sorry if this is rehashing old stuff to you folks. Yes, choice of screen fonts can partially mitigate the problem but characters will always be thin and fuzzy compared to their subpixel versions of the same font. Jerry > >> But already we should be able to find out why the pixmaps are too short. > > Yes hopefully, now I've tried 5.6.0-alpha Qt and it doesn't improve the > situation. > > Stephan > >> Also, the bearing computation will be completely wrong with bidi text. >> >> JMarc >
