Summary: | A shading displaced when rendering a re-distilled Altona-Testsuite_p2_S_x3.pdf | ||
---|---|---|---|
Product: | Ghostscript | Reporter: | leonardo <leonardo> |
Component: | Graphics Library | Assignee: | Robin Watts <robin.watts> |
Status: | RESOLVED FIXED | ||
Severity: | normal | CC: | alex, christinedelight.top85, shailesh.mistry |
Priority: | P4 | Keywords: | bountiable |
Version: | master | ||
Hardware: | PC | ||
OS: | Linux | ||
Customer: | Word Size: | --- | |
Attachments: |
out6665.pdf
decompr-out6665-.pdf cur1.ppm.zip |
Description
leonardo
2006-03-20 11:38:25 UTC
Created attachment 2114 [details]
out6665.pdf
Attaching a copy of the PDF file.
This bug looks similar to 688407, but actually it's another problem. That one
was about pdfwrite, this one is about raster devices.
Oops, the bug description lists a wrong test case. It should be Altona- Testsuite_p2_S_x3.pdf instead "STEUER-RollingMesh 3(Final).ai". Created attachment 2115 [details]
decompr-out6665-.pdf
A simplified test file : unrelated graphics removed, unrelated clip path
removed, colors changed for a better visibility.
The verticall line appears due to 0.5 .setfilladjust in gs_init.ps for small resolutions. The left half of the circle is painted with clippig [10734.1, 11379.1) by X, and the right half is painted with the clipping [11379.1, 12024.2). They round to [1073, 1138) and [137, 1203) device pixels. In same time, the right half is painted with filling a path, which covers [11379, 11970) by X. The letter applies a quoter pixel fill_adjust, and then rounds to [1138, 1197) device pixels. Thus the right half doesn't paint the pixel x=1137. A better result would appear if fill_adjust is equal to 0.5 pixels. This is another time when we detect an incorrect behavior with fill_adjust=0.5 . I think that fill_adjust=0.25 must apply to 'stroke' only, or stroking must implement a smarter logic, when fill_adjust is accounted in the line width. 'fill' operators must always use fill_adjust=0.5 . Passing the bug to Ray, who defined the fill_adjust logic. Please note a similar bug 688407 is important for a customer. Created attachment 2116 [details]
cur1.ppm.zip
The raster generated with -r72. A vertical line appears across the shading.
A shading with line problem. Likely similar to the shading line problem that Robin was just working on so assigning to him. Bug still reproducible in Ghostscript 9.03 This is fixed in v. 9.09 and higher. The bug can be closed. I agree with Alex's comment #8, and so I'm closing this bug. |