GNOME Bugzilla – Bug 547221
dvi backend of 2.23 incorrectly displays overllaping Type 1 font glyphs
Last modified: 2010-10-12 09:49:55 UTC
The dvi backend of evince 2.23 converts Type 1 fonts to PK, which it probably should not do, and then incorrectly displays the resulting glyphs if they overlap in the dvi file. Both xdvi and dvipng can display the Type 1 glyphs properly, probably because they don't gratuitously convert to PK fonts. I'm linking to a screenshot that is using a commercial font to make the problem easy to spot [http://www.cs.umd.edu/~gaburici/Screenshot-evince.dvi.png]. With fonts which are less slanted and more letter-spaced it's harder to see what's going on.
Could you attach here the dvi document, please?
Created attachment 116377 [details] DVI
It works for me.
Meaning that the glyph white areas do not obscure the black areas? I'm using freetype 2.3.7. Does evince depend on anything else to render dvis?
Created attachment 116798 [details] Screenshot (In reply to comment #4) > Meaning that the glyph white areas do not obscure the black areas? Yes, this is what I see. > I'm using freetype 2.3.7. Does evince depend on anything else to render dvis? > freetype 2.3.5 (ubuntu) cairo 1.6.5 (from sources)
What you are seeing is the broken behavior! Note the E is partially hidden by the white bounding box of the X. This does not happen with xdvi or dvipng. Doesn't seem to matter, but I'm using: freetype-2.3.8-0.1.20080729cvs.fc9.i386 cairo-1.6.4-1.fc9.i386
(In reply to comment #6) > What you are seeing is the broken behavior! Note the E is partially hidden by > the white bounding box of the X. This does not happen with xdvi or dvipng. > oh, you are right. Confirming then. > Doesn't seem to matter, but I'm using: > freetype-2.3.8-0.1.20080729cvs.fc9.i386 > cairo-1.6.4-1.fc9.i386 >
I guess it's caused by this change: http://svn.gnome.org/viewvc/evince?view=revision&revision=3006 See the explanation on mailing list: http://mail.gnome.org/archives/evince-list/2008-April/msg00011.html
Hmm, are you sure? IMHO the problem is that evince resorts to bitmaps, which are probably not transparent, so cairo can't help but obscure what's behind them. OTOH, xdvi uses PostScript outlines, no bitmaps are generated.
(In reply to comment #9) > Hmm, are you sure? No, I'm not sure. I just guessed that was the problem because it's the only change we have made in the dvi backend lately and it's related to T1 fonts. > IMHO the problem is that evince resorts to bitmaps, which > are probably not transparent, so cairo can't help but obscure what's behind > them. OTOH, xdvi uses PostScript outlines, no bitmaps are generated. >
I think this is a duplicate of 494736. With the patch I added to that defect, this problem goes away on the attached DVI.
Thanks for the bug report. This particular bug has already been reported into our bug tracking system, but please feel free to report any further bugs you find. *** This bug has been marked as a duplicate of bug 494736 ***