2013/4/13 Khaled Hosny <[email protected]>: > On Sat, Apr 13, 2013 at 01:12:50PM +0200, Zdenek Wagner wrote: >> 2013/4/13 Khaled Hosny <[email protected]>: >> > On Wed, Apr 10, 2013 at 11:32:47AM +0200, Zdenek Wagner wrote: >> >> 2013/4/9 Khaled Hosny <[email protected]>: >> >> > On Tue, Apr 09, 2013 at 04:37:19PM +0200, Zdenek Wagner wrote: >> >> >> 2013/4/9 Khaled Hosny <[email protected]>: >> >> >> > Hi Zdenek, >> >> >> > >> >> >> > For all the failing tests I get the same output from Uniscribe as >> >> >> > from >> >> >> > HarfBuzz and the new XeTeX, as such I believe those are font bugs. >> >> >> > The >> >> >> > difference is probably because the font now contains version 2 Indic >> >> >> > features >> >> >> > which HarfBuzz will always use (there is no way to tell HarfBuzz to >> >> >> > use >> >> >> > the old features if the new ones are present). >> >> >> > >> >> >> Thanks for your effort. Steve White tried HTML version of these tests >> >> >> in Windows Vista and FreeFont worked there (he sent me the >> >> >> screenshot). Anyway, your build fails even with FreeFont 20120503 >> >> >> (distributed with TL2012) and if I remember well, it does not have >> >> >> version 2 features (but maybe my memory is wrong). >> >> > >> >> > I need his test files then. >> >> > >> >> I do not have his exact files but the attached file contains his >> >> screenshot. I have typed the text in HTML, you just have to add a CSS >> >> defining the font to be used. >> > >> > I'm not sure I follow, but the output in your screenshot is the exact >> > same output I get with XeTeX, what is the issue then? >> > >> It worked fine in TL'12, the PDF's that failed are in directory >> http://hroch486.icpf.cas.cz/xetex-test/0.9999.2-r2525/ (ie XeTeX >> 0.9999.2 and FreeFont rev 2525), see vatu.pdf in that directory. > > There are two things to compare against here: > > * the pre-0.9999 output (i.e. our patched ICU LE). > * the Uniscribe output. > > Now given the situation of complex script support in OpenType, whatever > Uniscribe does *is* the standard and if other renderers deviate form > it, they are wrong (unless the deviation is deliberate, which is not the > case here with AFAIK). > > So, what I'm saying is that 0.9999 output (i.e. HarfBuzz) is the same as > Uniscribe, so it is right and the old output was wrong. You were > implying that Uniscribe on Windows 8 was giving different result from > 0.9999 but the screenshots you gave me show otherwise. > > So in short, unless Uniscribe is giving you *different* results than > XeTeX 0.9999, the new behaviour is correct and the old one is wrong. > Sorry but saying that fonts used to work only due to equivalent lucky bugs in both ICU and Pango and all firefox versions distributed with CentOS and Fedora within the last 8 years (including the current version 20.0) sounds really strange to me. To be sure I have just tested a few private web pages relying on locally installed FreeFont as well as http://hroch486.icpf.cas.cz/freefont-devanagari/ containing the font on the server and everything works. Of course it would bother me if other shapers generated wrong output, I would like to prepare EPUB with Hindi texts.
Is there a possibility to obtain the output of my samples (preferably online) from Uniscribe if I send the text and the font? I would like to report just what is wrong, not to bother with tests that pass. I have only legal installation media of Windows 98 and Windows XP (but not installed anywhere because my old computers have crashed) and do not intend to buy newer versions of Windows. I plan to buy an iPad when I earn some money. > Regards, > Khaled > > > -------------------------------------------------- > Subscriptions, Archive, and List information, etc.: > http://tug.org/mailman/listinfo/xetex -- Zdeněk Wagner http://hroch486.icpf.cas.cz/wagner/ http://icebearsoft.euweb.cz -------------------------------------------------- Subscriptions, Archive, and List information, etc.: http://tug.org/mailman/listinfo/xetex
