Bug 690998

Summary: x11 depth 15 and depth 16 have visual artefacts compared to depth 24.
Product: GhostPCL Reporter: Hin-Tak Leung <htl10>
Component: PCL rasterAssignee: Chris Liddell (chrisl) <chris.liddell>
Status: CONFIRMED ---    
Severity: normal CC: christinedelight.top85, ghostpdl-bugs, michael.vrhel
Priority: P4    
Version: unspecified   
Hardware: PC   
OS: Linux   
Customer: Word Size: ---
Attachments: zip'ed, retest, of 8,15,16,24 bits, and their diffs. 7 files.

Description Hin-Tak Leung 2009-12-14 20:39:15 UTC
This is a follow up to bug 690328 - it is possible to test different X-server
bit depth quite conveniently with XvFb (the frame buffer X server, part of
recent Xorg) like this:

  Xvfb :1 -screen 0 800x1100x15 &
  DISPLAY=:1 pspcl6 -sDEVICE=x11 /home/Hin-Tak/Desktop/468-01.ps.pxl

and in a different console, inspect the invisible frame buffer by running a
screen-grab:

  xwd -display :1 -root -out depth15.xwd
  convert depth15.xwd depth15.png

and have :2, :3 at different depth.
The pxl data from bug 690972
http://bugs.ghostscript.com/attachment.cgi?id=5732&action=view shows different
look at in the top right left box.

png figures to follow.
Comment 1 Hin-Tak Leung 2009-12-14 20:40:21 UTC
Created attachment 5768 [details]
depth 15
Comment 2 Hin-Tak Leung 2009-12-14 20:40:56 UTC
Created attachment 5769 [details]
depth 16
Comment 3 Hin-Tak Leung 2009-12-14 20:45:39 UTC
Created attachment 5770 [details]
depth 24 for reference.

depth 24 for reference. Compare the top left box. This may be related to 
bug 690972 (pgmraw vs ppmraw output difference on the same file). depth 16
looks like the pgmraw defect; whereas depth 15 is yet different.

The original file when runs through pxlcolor works alright under 15/16/24, so
this appear to be a pxlmono<->color-depth interaction.
Comment 4 Hin-Tak Leung 2009-12-14 20:46:02 UTC
Comment on attachment 5768 [details]
depth 15

derived from private data
Comment 5 Hin-Tak Leung 2009-12-14 20:46:20 UTC
Comment on attachment 5769 [details]
depth 16

derived from private data
Comment 6 Hin-Tak Leung 2009-12-14 20:46:39 UTC
Comment on attachment 5770 [details]
depth 24 for reference.

derived from private data
Comment 7 Henry Stiles 2010-01-07 09:46:17 UTC
Hin-Tak what is the status of this bug?
Comment 8 Hin-Tak Leung 2010-01-11 20:19:22 UTC
still a problem (look no change) with ghostpdl r10595 + gs trunk r10592 . 

I have also a ghostpdl r10595 + gs icc branch r10596 build, which shows
interestingly different look at depth 15/16, and at depth 24 is different 
from gs trunk r10592.

Since ghostpdl built with gs icc branch r10596 behaves differently from gs
trunk, I'll add Michael in CC, as a ghostpdl icc-related regression matter.
Comment 9 Hin-Tak Leung 2010-01-11 20:20:41 UTC
meant to add Michael in the last comment regarding ghostpdl icc build behaving
differently from trunk build.
Comment 10 Hin-Tak Leung 2010-01-11 20:23:23 UTC
Created attachment 5863 [details]
depth 15 icc branch look

depth 15 icc branch screen capture.
Comment 11 Hin-Tak Leung 2010-01-11 20:24:12 UTC
Created attachment 5864 [details]
depth 16 icc branch x-server screen capture

depth 16 icc branch x-server screen capture
Comment 12 Hin-Tak Leung 2010-01-11 20:25:16 UTC
Created attachment 5865 [details]
depth 24 icc branch X-server screen capture

depth 24 icc branch X-server screen capture
Comment 13 Hin-Tak Leung 2010-01-11 20:25:37 UTC
Comment on attachment 5863 [details]
depth 15 icc branch look

derived from private data
Comment 14 Hin-Tak Leung 2010-01-11 20:25:57 UTC
Comment on attachment 5864 [details]
depth 16 icc branch x-server screen capture

derived from private data.
Comment 15 Hin-Tak Leung 2010-01-11 20:26:18 UTC
Comment on attachment 5865 [details]
depth 24 icc branch X-server screen capture

derived from private data
Comment 16 Henry Stiles 2010-02-02 14:35:18 UTC
What is the status of this Hin-Tak?  Should it be reassigned to Michael?
Comment 17 Hin-Tak Leung 2010-02-08 23:49:17 UTC
Tried with icc branch r10714-r10720 and trunk-r10714-r10710 and problem as before.
Re-assign to Michael on Henry's suggestion.

I'll be happy to test patches or even get my hands dirty look at specific area
of code. (this is ghostpdl processing pxl to render to 15/16-bit color vs
24-bit, and I don't have much clue where the problem might be even at a guess). 
Comment 18 Hin-Tak Leung 2014-07-25 14:10:07 UTC
Created attachment 11080 [details]
zip'ed, retest, of 8,15,16,24 bits, and their diffs. 7 files.

zip'ed, retest, of 8,15,16,24 bits, and their diffs. 7 files in png.

Just re-visiting - still a problem; also note that although 8-bit is closest to 24- result, they still differ enough.