forwarded 522675 http://bugzilla.gnome.org/show_bug.cgi?id=578336
Am Sonntag, den 05.04.2009, 21:22 +0200 schrieb Marc Lehmann: > Package: libpango1.0-0 > Version: 1.24.0-2 > Severity: normal > > > after upgrading to libgpango 1.24, my applications started to crash like > this: > > (process:19388): Pango-WARNING **: failed to choose a font, expect ugly > output. engine-type='PangoRenderFc', script='latin' > > (process:19388): Pango-WARNING **: failed to choose a font, expect ugly > output. engine-type='PangoRenderFc', script='common' > > (process:19388): Pango-WARNING **: failed to choose a font, expect ugly > output. engine-type='PangoRenderFc', script='hiragana' > ./deliantra: line 6: 19388 Segmentation fault (core dumped) > > The reason is coimmit 9305b9a9995e84ace3818a90346820ef8c47a3ce to pango: > > [pangofc] Add a "fontmap" property to PangoFcFont > > Gecko uses its own PangoFcFontMap subclass with its own > PangoFontSet. Previously we were setting font->fontmap in our > own private PangoFcFontSet. Now it's up to the PangoFcFont > subclass to set it when creating the new font object. > > applications subclassing the public pangofc api that did work properly > will now fail. The problem is that doing what the patch suggests works > with 1.24, but fails with 1.22: > > > Pango:ERROR:/tmp/buildd/pango1.0-1.22.4/pango/pangofc-fontmap.c:559:pango_fc_font_map_add: > assertion failed: (fcfont->fontmap == NULL) > > So, while adding: > > #if PANGO_VERSION_CHECK (1, 23, 1) > PANGO_FC_FONT (font)->fontmap = fontmap; > #endif > > makes the program work, you need a recompile betwene 1.22 and 1.24, as > 1.22 checks that the fontmap is 0, and 1.24 requires it to be non-null. > > This makes it hard (nothing short of a run-time check possible) to create > binaries that work with 1.22 and 1.24. > > (My understanding of this might be wrong, but such a big incompatible > api change deserves at leats a better error message, especially when > older versions of pango explicitly made setting the fontmap illegal by an > assertions). Thanks, I've forwarded this bug upstream. Let's see what they say about it... it'd be really bad if there was such an ABI breakage in 1.24.
signature.asc
Description: Dies ist ein digital signierter Nachrichtenteil