On 6/19/07, Drew Parsons <dparsons@debian.org> wrote:
OK, you're using the stable etch versions. One possibility is that the matter got addressed and fixed in the following versions that we're now testing in unstable. I've reproduced the xfd box size difference, but let's try for reproduction of the latin sample difference.
If the character metrics in xfd windows differ then I'd say that the bug is present in newer Xorg packages.
> Another two screenshots contain text samples made by Dejavu and Arial > fonts. The upper sample is produced using 8-bit font encoding > (iso8859-1 for dejavu and koi8-r for arial - it contains some > cyrillic). The problem is that the samples look differently, which is > unacceptable in my view. > > I've also attached magnified samples for which the differences are > highlighted by thing red strokes. OK, that illustrates the problem well. Let's focus on reproducing the dejavu sample difference you've shown here. Can you describe precisely how you generated these dejavu samples?
I used the attaches Tcl/Tk script. Tcl/Tk 8.4.12, which is shipped with etch (and with sid also) uses old X fonts mechanism, which contains the bug (I think).
Alternatively, I can generate text samples from gfontview using both "gfontview" and "LANG=en_AU gfontview" (my default locale is UTF-8, so I expect the former to be using the iso10646-1 version of the font, while the latter should be using iso8859-1. ).
I'm pretty sure that gfontview uses Xft for font rendering. It's a different method, for which I didn't see this bug. So, to reproduce the bug, you have to use some other software. xfontsel comes to mind (with -sample, -sample16, -sampleUCS options). I select the dejavu font,
enter some string "In girum consum" into the Test String then click 'Make String'. I can magnify the resulting sample using xmag. In my case the horizontal spacing appears to be the same. It might be useful if you could try this method to see if your differences can be reproduced this way. A third technique to get samples is to start up abiword (applying LANG as needed).
Abiword is based on GTK+2 too, so it can't show the bug too.
I guess for your system, you'd use LANG=ru_RU.KOI8-R, LANG=ru_RU.UTF-8, maybe LANG=ru_RU.ISO-8859-1 (which doesn't really make sense) or LANG=en_US.ISO-8859-1 or maybe LANG=C. But anyway, tell us what you did to generate your dejavu text samples. Oh, tell us what your default locale is too. "env | grep LANG". Mine is en_AU.UTF-8.
My default locale is ru_RU.CP1251. The attached script uses explicit font names for both labels. -- Sergei Golovan
Attachment:
sample.tcl
Description: Tcl script