A vertical line dissappears in the box 014-01 when rendering re-discrtilled 014- 01.ps with HEAD. Adobe paints it. I guess the 'arct' may be not conforming in the PDF interpreter. Same with 015-01.ps .
Created attachment 896 [details] 014-01.pdf The test file generated with gswin32c.exe -dNOPAUSE -dNOOUTERSAVE -dBATCH -sDEVICE=pdfwrite -sOutputFile=014-01.pdf -c false 0 startjob pop -f - <comparefiles/014-01.ps
Reported by Julia as a regression from Dec 09, 2003. Reproduced by Igor with today's CVS HEAD
To reproduce the problem : gswin32c.exe-r300 -dNOPAUSE -sDEVICE=ppmraw -sOutputFile=z:\cur.ppm 014- 01.pdf -c quit
This shows different drop outs at different resolutions. At -r120 the bottom part of a curve is missing. At -r72 a vertical segment is missing.
At 120 dpi, doing 0.5 .setfilladjust makes the horizontal dropout on the curve disappear: gs -r120 -c 0.5 .setfilladjust 014-01.ps The vertical lines missing at 72 dpi do not appear to be affected by increasing the .setfilladjust setting. Most likely this is a result of fill coordinates in those vertical 'lines' being rounded to a 0 width fill (same X coordinates going up and down). Raph feels that this is an undefined area of the Adobe PS spec. One thing that is noted is that we should set the fill adjust to 0.5 if we are using GraphicsAlphaBits > 1 so that anti-aliasing will work better. This can be seen at 120 dpi, although the anti-aliased rendering still shows a dropout on the curve (due to a low gray intensity being rounded down rather than up ?)
Created attachment 5478 [details] 687677-tiffs.zip Tiff files from Acrobat at 72, 100 and 120 dpi.
Reassigning path and fill problems to Robin Watts.
Reassigning to new email address.
Bug still reproducible in Ghostscript 9.03
The problem is still in the current master branch.