Bug 353454 – app-text/ghostscript-gpl-9 problem with RGBW color model

• Status: RESOLVED FIXED
• Product: Gentoo Linux

[Description Oldrich Jedlicka 2011-02-01 21:23:08 UTC]

This is a continuation of a color model problem as reported in bug #341207.

The output isn't right due to arithmetic overflow in hplip, but is primarilly caused by ghostscript-gpl bug.

Patch for hplip to have at least some color printing is in

  https://bugs.launchpad.net/hplip/+bug/695962

[Comment 1 Oldrich Jedlicka 2011-02-01 21:24:06 UTC]

Created attachment 261289 [details]
Problem with color model as printed by hplip

[Comment 2 Oldrich Jedlicka 2011-02-01 21:31:08 UTC]

Just to finish the description. The problem is with printing. The ghostscript-gpl with the CUPS driver provides wrong raster file, which is then passed to hplip. This results in arithmetic overflows and wrongly printed colors - see attachment.

[Comment 3 Timo Gurr (RETIRED) 2011-02-02 16:11:17 UTC]

Created attachment 261360 [details]
hplip-testpage.png

HPLIP Testpage printed using the following versions:

app-text/ghostscript-gpl-9.00-r1
net-print/cups-1.4.6
net-print/hplip-3.11.1-r2

Seems to look ok here.

[Comment 4 Timo Gurr (RETIRED) 2011-02-02 16:23:37 UTC]

Not sure about the RGB option though, the standard seems to be sRGB for Color LaserJet printers, the options offered by CUPS/the hplip ppd are:

* Standard (sRGB)
* Luminous
* Photo
* Photo (Adobe RGB 1998)
* None

[Comment 5 Oldrich Jedlicka 2011-02-02 21:59:32 UTC]

Ok, I will check the latest packages too (tomorrow).

[Comment 6 Oldrich Jedlicka 2011-02-03 05:44:06 UTC]

(In reply to comment #5)
> Ok, I will check the latest packages too (tomorrow).

I had a quick look and the raster file looks much better with ghostscript-9.00-r1. I will have a look at the applied patches to see more. Still the Color Wheel and the HP logo is broken, but the rest looks (at the first sight) how it should be - RGB is part is white and the W part contains the drawing (from the RGBW color model). I will attach my test page print-out in the evening too.

[Comment 7 Timo Gurr (RETIRED) 2011-02-17 13:43:01 UTC]

I could reproduce the issue with an ink printer connected via usb. Attached are the results using different ghostscript versions, latest one looking the worst..

[Comment 8 Timo Gurr (RETIRED) 2011-02-17 13:43:52 UTC]

Created attachment 262829 [details]
gentoo-gs-9.00.png

[Comment 9 Timo Gurr (RETIRED) 2011-02-17 13:44:11 UTC]

Created attachment 262831 [details]
gentoo-gs-9.00-r1.png

[Comment 10 Timo Gurr (RETIRED) 2011-02-17 13:44:28 UTC]

Created attachment 262833 [details]
gentoo-gs-9.01.png

[Comment 11 Oldrich Jedlicka 2011-02-17 13:55:21 UTC]

Hm, there wasn't any update from Ghostscript developpers in the referenced bug report, so I'm not surprised it isn't fixed in 9.01.

[Comment 12 Timo Gurr (RETIRED) 2011-02-17 14:09:37 UTC]

(In reply to comment #11)
> Hm, there wasn't any update from Ghostscript developpers in the referenced bug
> report, so I'm not surprised it isn't fixed in 9.01.

Indeed, but strange thing is that my testpage with 9.01 looks like yours with 9.00, while you got better results with 9.00-r1. -r1 just had upstream patches applied which are now in 9.01. I'll commit 9.01 later today, 9.xx won't go stable until this issue is resolved.

[Comment 13 Alec Meyers 2011-02-17 16:29:02 UTC]

Timo, could you *please* mask the non-working versions of Ghostscript?

[Comment 14 Alec Meyers 2011-02-17 16:34:46 UTC]

Created attachment 262845 [details]
black-on-white graphviz diagram (ghostscript-9.00)

Here's a printout I found, if it seems like not much of a problem...

[Comment 15 Oldrich Jedlicka 2011-02-17 21:47:12 UTC]

(In reply to comment #14)
> Created an attachment (id=262845) [details]
> black-on-white graphviz diagram (ghostscript-9.00)
> 
> Here's a printout I found, if it seems like not much of a problem...

This has been fixed in 9.00-r1/9.01 I think. Version 9.00 definitely wasn't able to render images correctly. This is clearly visible on the HP logo (image) on the test page, while the rest of the page (vector drawings) looks good.

When you tried to print scanned pages from PDF, print image from Gimp or _rasterized_ drawings from Inkscape, you had somehow inverted colors. This should be fixed.

[Comment 16 Christian Loosli 2011-03-08 21:29:53 UTC]

Sorry to tell you, but this is definitely not fixed in ghostscript 9.01. 

I have a HP B110 printer here, and I tried hplip 3.10.9-r1 and 3.11.1-r2. As soon as I use ghostscript 9.* the colors are wrong if I print a testpage via hplip or just any picture. As soon as I downgrade to ghostscript 8.71-r6 the problems are gone. The USE-Flags of hplip (hpjis, hpcups) didn't seem to affect this. 

Just tell me if you need more information regarding versions in use or similar. 

Kind regards, 

Christian

[Comment 17 Oldrich Jedlicka 2011-03-09 06:03:49 UTC]

(In reply to comment #16)
> Sorry to tell you, but this is definitely not fixed in ghostscript 9.01. 

I meant only the part with inverted images. There are still problems with colors, though (see attachment 262833 [details] that nicely shows the problem).

[Comment 18 Oldrich Jedlicka 2011-08-01 19:59:09 UTC]

Latest ghostscript-gpl git should have the issue fixed. I'm trying to get it running, currently without much success, though...

[Comment 19 Timo Gurr (RETIRED) 2011-08-10 11:06:08 UTC]

Please retest with ghostscript-gpl-9.04.

(In the meantime using ghostscript-gpl-9.02 and cups-1.5_rc also worked for me printing a hplip testpage showing the right colors.)

[Comment 20 Alexandre Ferreira 2011-08-14 02:55:00 UTC]

Seeing same problem with ghostscript-gpl-9.04, cups-1.4.6-r21 and hplip-3.11.7. My HP Deskjet 3900 only prints in black and white.

[Comment 21 Alexandre Ferreira 2011-08-14 06:20:55 UTC]

It works with ghostscript-gpl-9.02, cups-1.4.6-r21 and hplip-3.11.7

[Comment 22 Oldrich Jedlicka 2011-08-15 05:57:37 UTC]

I've reopened the upstream bug and added information about the BW problem.

[Comment 23 Oldrich Jedlicka 2011-08-15 06:02:14 UTC]

The 9.04 problem is printed image in grayscale - to be precise. Only the W part of the RGBW image contains image data.

[Comment 24 Oldrich Jedlicka 2011-08-16 05:55:26 UTC]

The problem has been fixed in the current Git code (master branch).

[Comment 25 Timo Gurr (RETIRED) 2011-08-16 21:07:19 UTC]

(In reply to comment #24)
> The problem has been fixed in the current Git code (master branch).

Fixed in ghostscript-gpl-9.04-r1. Huge thanks for taking care of this bug upstream and investing so much time in finding the root cause, greatly appreciated!

[Comment 26 Oldrich Jedlicka 2011-08-17 05:31:45 UTC]

(In reply to comment #25)
> (In reply to comment #24)
> > The problem has been fixed in the current Git code (master branch).
> 
> Fixed in ghostscript-gpl-9.04-r1. Huge thanks for taking care of this bug
> upstream and investing so much time in finding the root cause, greatly
> appreciated!

Version ghostscript-gpl-9.04-r1 generates the same output as Git master. Thanks!

[Comment 27 Andrew Savchenko 2011-08-17 17:24:25 UTC]

ghostscript-gpl-9.04-r1 fails to build for me on ~x86 without SSE2 support, I opened new bug 379597 for this.

Version 9.02 builds well.

[Comment 28 Oldrich Jedlicka 2011-08-22 18:38:48 UTC]

There is one more commit upstream to fix color model changes.