Gnumeric strange colors (purple charts) on bigendian

Bug #109204 reported by mambro
6
Affects Status Importance Assigned to Milestone
Gnumeric
Fix Released
High
gnumeric (Ubuntu)
Invalid
Undecided
Unassigned
Nominated for Feisty by sam tygier
goffice (Ubuntu)
Fix Released
Undecided
Unassigned
Nominated for Feisty by sam tygier

Bug Description

This bug causes colours to be miscalculated when displaying charts in a gnumeric spreadsheet on powerpc (bigendian) computers.

It was fixed upstream by changing some functions for colour calculations in goffice/utils/go-image.c, these are within a
#if G_BYTE_ORDER == G_LITTLE_ENDIAN
so i386 should be unaffected.

The fix has been committed upstream and included in the 0.4 release of goffice
http://bugzilla.gnome.org/show_bug.cgi?id=432532
http://svn.gnome.org/viewcvs/goffice?view=revision&revision=1730

A patch has been attached to this bug report.

The bug can be reproduced by:
opening Gnumeric
Insert-> Chart
select XY chart type, press insert
drag an area for the chart to be drawn on the spread sheet.

=======
old description
Binary package hint: gnumeric

I have ubuntu feisty 7.04 on an ibook G4 (ppc version).
When i try to make a chart it appears purple. Same thing with the windows "About gnumeric".

This is a screenshot http://img100.imageshack.us/img100/6336/schermatave3.png

I had the same problem with gnumeric 1.7.* with Gentoo on PPC.

Related branches

mambro (mambro87)
description: updated
Revision history for this message
sam tygier (samtygier) wrote : Re: Gnumeric strange colors (purple charts)

i am also seeing this on a powerbook G4

Changed in gnumeric:
status: Unconfirmed → Confirmed
Changed in gnumeric:
status: Unknown → Unconfirmed
Revision history for this message
mambro (mambro87) wrote :

Why ubuntu has 1.7.x version that is a development branch and not 1.6.x that is considered stable?

Revision history for this message
sam tygier (samtygier) wrote :

jean from gnumeric has tracked this down to the cairo_to_pixbuf() function in goffice/utils/go-image.c

i'll try to make a patch for ubuntu

Changed in gnumeric:
assignee: nobody → samtygier
Revision history for this message
sam tygier (samtygier) wrote :

the upstream patch is at http://bugzilla.gnome.org/attachment.cgi?id=87150 it only touches 4 lines.

it applies cleanly to the ubuntu goffice 0.3.7-0ubuntu1 package. i have built the package and tested it on powerpc

http://tygier.co.uk/pub/libgoffice-0-3_0.3.7-0ubuntu1_powerpc.deb

the patch solves the issue for me, and i can't see any regressions.

Changed in goffice:
status: Confirmed → Fix Committed
Revision history for this message
mambro (mambro87) wrote :

The package works also for me

Changed in gnumeric:
status: Unconfirmed → Fix Released
Revision history for this message
sam tygier (samtygier) wrote :

i have built a patched version on x86, and could see not regression in graph drawing.

i.e. the patch fixes big endian, without breaking little endian

(i dont think this should be assigned to me any more, as i dont think there is any more work i can do.)

Changed in goffice:
assignee: samtygier → nobody
Revision history for this message
mambro (mambro87) wrote :

You need some further tests? if you need I do something ask me.
I ask you this because i don't see the package in the official tree yet..

Revision history for this message
casfindad (caseyfmail-linux) wrote : Update Manager returns bug

Sam's patch also worked for my Feisty install on a Powerbook G4. Thank you Sam! Upgrading Gnumeric was one of the main reasons I upgraded to Feisty. (Error bars were broken in the Edgy release.)

I do have my Update Manager nagging me to "update" libgoffice-0-3. Thinking your patch was already in the repositories, I updated and the bug came back. Re-installing your patched deb fixed it again though.

I hope the patched version goes into the repositories soon.

Revision history for this message
sam tygier (samtygier) wrote :
Revision history for this message
mambro (mambro87) wrote :

They have realased goffice 0.4.0 and gnumeric 1.7.10 that solve the bug
http://www.gnome.org/projects/gnumeric/announcements/1.7/gnumeric-1.7.10.html

I know that ubuntu policy say not to change software version in the same release but gnumeric 1.7.x are a development versions so it's normal that they aren't perfect..

Revision history for this message
casfindad (caseyfmail-linux) wrote :

That's good news about this bug being fixed in the next release, but can't Sam's fix be incorporated into the current libgoffice-0-3 for those of us who don't want to update goffice and gnumeric to a version that may be less stable in feisty? Update Manager keeps nagging me to update libgoffice, and I don't know if it wants to replace my patched libgoffice from Sam with the broken one in the repos. Last time I agreed to let libgoffice update, the repo version had the bug and I had to re-install Sam's version.

Revision history for this message
mambro (mambro87) wrote :

Someone can tell me why this fix is not yet avaiable?!?!

Revision history for this message
sam tygier (samtygier) wrote :

SRU proposal

Bug #109204

This bug causes colours to be miscalculated when displaying charts in a gnumeric spreadsheet on powerpc (bigendian) computers.

It was fixed upstream by changing some functions for colour calculations in goffice/utils/go-image.c, these are within a
#if G_BYTE_ORDER == G_LITTLE_ENDIAN
so i386 should be unaffected.

The fix has been committed upstream and included in the 0.4 release of goffice
http://bugzilla.gnome.org/show_bug.cgi?id=432532
http://svn.gnome.org/viewcvs/goffice?view=revision&revision=1730

A patch has been attached to this bug report.

The bug can be reproduced by:
opening Gnumeric
Insert-> Chart
select XY chart type, press insert
drag an area for the chart to be drawn on the spread sheet.

This patch fixes the bug on powerpc, and has been tested on powerpc and i386. No regressions have been reported

description: updated
Revision history for this message
Martin Pitt (pitti) wrote :

Closing gnumeric task, problem is in goffice

Changed in gnumeric:
status: Unconfirmed → Rejected
Revision history for this message
Martin Pitt (pitti) wrote :

This does not fit the SRU criteria at all, it's neither a really serious regression nor data loss. Also, it needs to be fixed in gutsy first (the bug task is still open for gutsy).

This would be an ideal candidate for a backport once it's fixed in gutsy.

Revision history for this message
mambro (mambro87) wrote :

How can you say that isn't a serious regression? Gnumeric means charts for most of the people. If i can't make charts I can't use gnumeric. It's more that serious!

Revision history for this message
mambro (mambro87) wrote :

How can you say that isn't a serious regression? Gnumeric means charts for most of the people. If i can't make charts I can't use gnumeric. It's more than serious!

Revision history for this message
sam tygier (samtygier) wrote :
Revision history for this message
mambro (mambro87) wrote :

I know you made a working package. I use it and I thank you. But Ubuntu is "linux for human beings" so most of the people don't read the forum and haven't installed your package. Ubuntu should work out of the box but, right now, gnumeric doesn't work on ppc.

Revision history for this message
Gauvain Pocentek (gpocentek) wrote :

goffice (0.4.0-1ubuntu1) gutsy; urgency=low

  * Merge with debian experimental
    - keep the gtk/gnome multibuild
    - keep calling intltool-update -p
  * Closes LP: #89879, LP: #109204, LP: #59529

goffice (0.4.0-1) experimental; urgency=low

  * New upstream development release.
  * Upstream fixes include goffice/utils/go-image.c: (cairo_to_pixbuf): fixed
    conversion on bigendian machines [gnome bugzilla #432532].
    (Closes: #411628)
  * [debian/*] Adapt to 0.4 version.

 -- Gauvain Pocentek <email address hidden> Mon, 4 Jun 2007 13:52:32 +0200

Changed in goffice:
status: Fix Committed → Fix Released
Changed in gnumeric:
importance: Unknown → High
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.