Bug 688588

Summary:	pattern fill not displayed with antialias, or in tiff
Product:	Ghostscript	Reporter:	Volker Kuhlmann <list0570>
Component:	Graphics Library	Assignee:	Ralph Giles <ralph.giles>
Status:	RESOLVED FIXED
Severity:	normal
Priority:	P3
Version:	8.53
Hardware:	PC
OS:	Linux
Customer:		Word Size:	---
Bug Depends on:	688933
Bug Blocks:
Attachments:	EPS test file which isn't handled correctly.

Description Volker Kuhlmann 2006-03-08 13:45:34 UTC

The attached .eps shows a 4-corner polygon with a cross-hatch fill. The  
cross-hatch is shown with -sDEVICE=x11, but not with  
  
-sDEVICE=x11 -dTextAlphaBits=4 -dGraphicsAlphaBits=4 -dMaxBitmap=50000000  
  
or -sDEVICE=tiff24nc. The polygon is always shown.  
  
AFPL Ghostscript 8.53 (2005-10-20)  
  
Is this a (longer standing) bug, or an unimplemented feature? When converting  
the .eps to .pdf with gs 8.53 ps2pdf14, both xpdf and acroread show the  
cross-hatch as expected. gs 8.53 has the same problems with the .pdf as with 
the .eps.

Comment 1 Volker Kuhlmann 2006-03-08 13:48:19 UTC

Created attachment 2085 [details]
EPS test file which isn't handled correctly.

Cross-hatch fill not shown with x11 device and antialias, or in tiff24nc.

Comment 2 Volker Kuhlmann 2006-03-09 17:38:32 UTC

Also see bug #688580

Comment 3 Dan Coby 2006-03-09 19:32:04 UTC

Verified as reported with 8.53.  Also apparent with the display device.

With SVN head abd the following command line:

bin/gswin32c -dGraphicsAlphaBits=4 ../test_files/688588.ps -c quit

I get the following:

AFPL Ghostscript CVS PRE-RELEASE 8.54 (2005-10-20)
Copyright (C) 2005 artofcode LLC, Benicia, CA.  All rights reserved.
This software comes with NO WARRANTY: see the file PUBLIC for details.
AFPL Ghostscript CVS PRE-RELEASE 8.54: Unrecoverable error, exit code 255

The error is coming from:

c:\_\artifex\gs\src\gxpcmap.c(361): Returning error -100

Which says:

return_error(gs_error_Fatal); /* can't happen *

Guess the comment is wrong.

Comment 4 Raph Levien 2006-03-15 09:44:40 UTC

We do not currently implement pattern tile accumulation devices with destination
alpha. The heart of this work would be implement standard Porter-Duff "over"
rules in the copy_alpha method. This may help with this bug, but it looks like
the main thing is filling in some unhandled cases.

Comment 5 Raph Levien 2007-01-10 11:08:52 UTC

By turning on the repeated-rendering path of bug #688826, as opposed to the pattern accumulator, we 
will likely get correct rendering even without implementing porter-duff alpha compositing in the pattern 
logic.

Comment 6 Ray Johnston 2009-01-07 11:09:09 UTC

This appears to have been fixed -- at least it doesn't fail and the output
looks reasonable with the command line:

gswin32c -dEPSFitPage -dGraphicsAlphaBits=4 -r100 bug_688588.ps