I wonder whether FcDefaultSubstitute() should better be called before FcConfigSubstitute(). Why? I tried to use a rule like the following to replace the scalable "Nimbus Sans L" with the "Adobe Helvetica" bitmap font for pixel sizes between 8 and 20: <match target="pattern"> <test name="family"> <string>Nimbus Sans L</string> </test> <test name="pixelsize" compare="more_eq"> <double>8</double> </test> <test name="pixelsize" compare="less_eq"> <double>20</double> </test> <edit name="family" mode="prepend" binding="same"> <string>Adobe Helvetica</string> </edit> <edit name="foundry" mode="prepend" binding="same"> <string>Adobe</string> </edit> </match> As "Nimbus Sans L" is a Helvetica clone, the "Adobe Helvetica" bitmap font should be a reasonably good replacement at small sizes for people who absolutely insist on using bitmaps. Although I don't want to use that myself and don't want to make it the default on SuSE Linux either, I explored whether such a rule is possible to be able to give users who greatly prefer bitmaps in small sizes an easy way to get mostly bitmaps on their desktop for small sizes and anti-aliasing for big sizes. Similar like it can be done with embedded bitmaps for CJK fonts. Quite a few users requested something like this. While testing the above rule, I found that testing for "pixelsize" when matching "pattern" doesn't work when only "size" was requested in the pattern by the application. For example when calling xfd -fa "Nimbus Sans L:size=10" the above rules will not match, even if pixelsize=size/72*dpi is within range tested. The reason is that FcDefaultSubstitute is called *after* FcConfigSubstitute in libXft. Therefore, pixelsize is not yet set while the above rules are applied. This makes it impossible to use any rules matching on pixelsize in the pattern, one never knows whether an application will request pixels like xfd -fa "Nimbus Sans L:pixelsize=16" or point. Using "size" instead of "pixelsize" in the above rules doesn't appear to make much sense either, then it will fail if the application requests a pattern with only "pixelsize" and no "size". And, rules which are meant to fix rendering issues at small pixelsizes should match on "pixelsize" and not on "size" because it depends on the display of course how many pixels a certain point size has. That's why the rules for the "Bitstream Vera" fonts in /etc/fonts/fonts.conf match on "pixelsize", e.g.: <match target="font"> <test name="family"> <string>Bitstream Vera Serif</string> </test> <test name="pixelsize" compare="less"> <double>7.5</double> </test> <edit name="hinting"> <bool>false</bool> </edit> </match> But these match on the target "font", and as explained above, this doesn't work reliably when matching on the target "pattern". Shouldn't it be possible to match reliably on "pixelsize" also with target pattern? Just for testing, I applied the two attached patches to call FcDefaultSubstitute(). This makes the above rule work for both xfd -fa "Nimbus Sans L:size=12:dpi=112" xfd -fa "Nimbus Sans L:pixelsize=18.6667:dpi=112" *But*, if I omit "dpi=112" (which is the correct dpi for my display), FcDefaultSubstitute() makes dpi default to "75" which results in wrong conversion from point to pixel for my display. I.e. just calling FcDefaultSubstitute() is probably also not the right solution. I think somehow "pixelsize" should be calculated from "size" using the correct dpi value and added to the pattern before the matching rules from the config files are applied to the pattern. But how to do this without any nasty side effects? -------------- next part -------------- A non-text attachment was scrubbed... Name: fc-match.patch Type: text/x-patch Size: 569 bytes Desc: not available Url : http://freedesktop.org/pipermail/fontconfig/attachments/20040404/23875026/fc-match.bin -------------- next part -------------- A non-text attachment was scrubbed... Name: p_fc-default-substitute-first.diff Type: text/x-patch Size: 468 bytes Desc: not available Url : http://freedesktop.org/pipermail/fontconfig/attachments/20040404/23875026/p_fc-default-substitute-first.bin -------------- next part -------------- -- Mike FABIAN <mfabian@xxxxxxx> http://www.suse.de/~mfabian 睡眠不足はいい仕事の敵だ。