Bug 157607

Summary: Firefox loses parts of the configuration on update
Product: [openSUSE] SUSE Linux 10.1 Reporter: Reinhard Max <max>
Component: FirefoxAssignee: E-mail List <bnc-team-mozilla>
Status: RESOLVED NORESPONSE QA Contact: E-mail List <qa-bugs>
Severity: Normal    
Priority: P5 - None CC: asklein, axel.braun, suse-beta, vetter
Version: RC 1   
Target Milestone: ---   
Hardware: Other   
OS: Other   
Whiteboard:
Found By: Development Services Priority:
Business Priority: Blocker: ---
Marketing QA Status: --- IT Deployment: ---

Description Reinhard Max 2006-03-13 15:28:39 UTC
It happened to me several times during the 10.1 beta phase, that after updating Firefox forgot parts of the configuration. The most obvious one I noticed was that opening a URL by pasting it into the main window with the middle mouse button didn't work anymore. Now, after the update to beta7 it looks like it has also forgotten that I want it to open a new tab when a link says it wants to open a new window.
Comment 1 Wolfgang Rosenauer 2006-03-14 05:20:14 UTC
That's another strange thing. I update my Firefox often as you can imagine and never saw such a behaviour.
The first thing you mention should only happen by setting middlemouse.contentLoadURL to false but the software default if nothing is set is true.
Do you still have a "broken" profile where you can check your prefs.js?
Comment 2 Reinhard Max 2006-03-14 09:49:23 UTC
No broken profile left, but when I look for middlemouse.contentLoadURL in about:config, it indicates that the default is "false", and it seems to be falling back to that default on some (not all) updates, even when the version number of FireFox hasn't changed.
Comment 3 Wolfgang Rosenauer 2006-03-14 10:23:58 UTC
that's strange. My default is "true". Could you please grep in /usr/lib/firefox
for contentLoadURL?
Comment 4 Reinhard Max 2006-03-14 10:52:54 UTC
Did so, and some further inspection of the result brought up this:
$ grep -i middle /usr/lib/firefox/greprefs/all.js
// Middle-mouse handling
pref("middlemouse.paste", false);
pref("middlemouse.openNewWindow", true);
pref("middlemouse.contentLoadURL", false);
pref("middlemouse.scrollbarPosition", false);
// Middle-mouse handling
pref("middlemouse.paste", true);
pref("middlemouse.contentLoadURL", true);
pref("middlemouse.openNewWindow", true);
pref("middlemouse.scrollbarPosition", true);
Comment 5 Wolfgang Rosenauer 2006-03-14 11:05:01 UTC
That's the same for me. The reason is that the file gets "compiled" with platform ifdefs and includes a default first and the Linux ifdef afterwards.
Maybe somehow/sometimes this double setting confuses the pref system.
(although I never saw this, this might happen under special circumstances)
I'll see what can be done.
Comment 6 Reinhard Max 2006-03-14 12:43:59 UTC
Could it be that something like a syntax error makes Firefox silently stop parsing the file after it has seen the generic version of middlemouse.contentLoadURL, but before it gets to see the Linux-speciffic one?
Comment 7 Wolfgang Rosenauer 2006-03-14 18:47:37 UTC
A syntax error could cause this I think. But I guess you didn't modify your global prefs?
Comment 8 Robert O'Callahan 2006-03-15 02:42:40 UTC
Can you attach your all.js to this bug?

If the default in all.js was changing, that would explain this bug. When it changes to 'true' we'd wipe out your preference in prefs.js because it is the default. Then when it changes to 'false' you'd lose your preference.
Comment 9 Wolfgang Rosenauer 2006-03-15 04:39:06 UTC
(In reply to comment #8)
> If the default in all.js was changing, that would explain this bug. When it
> changes to 'true' we'd wipe out your preference in prefs.js because it is the
> default. Then when it changes to 'false' you'd lose your preference.

That's what I've thought before in the other bug mentioning this. But the contentLoadURL pref was true for Linux at least since 1.0.x up to our 1.5.0.1 package AFAIK.
Comment 10 Reinhard Max 2006-04-13 17:46:14 UTC
FWIW, it happened again when updating from (IIRC) beta6 to RC1.
Comment 11 Axel Braun 2006-04-25 18:30:20 UTC
...and on 10.0, when updating to FF 1.5.0.2
Comment 12 Reinhard Max 2006-04-26 07:41:03 UTC
It didn't happen when I updated from 10.1 RC1 to RC2.
Comment 13 Reinhard Max 2006-04-28 12:20:42 UTC
... and it happened again when I updated from RC2 to RC3 :(
Comment 14 Reinhard Max 2006-06-21 16:47:30 UTC
... and again when I tested SLES10 (Firefox 1.5.0.4) and then again when switching back to 10.1 (Firefox 1.5.0.3).
Comment 15 Axel Braun 2006-08-05 11:25:31 UTC
woa....1.5.0.4, 1.5.0.5, 1.5.0.6....every update the same story.....is this bug persistent?
Comment 16 Mark Gordon 2007-11-02 19:06:49 UTC
Any updates on how the upgrade to 2.0.0.* went? Is this still an issue?
Comment 17 Stephan Kulow 2007-11-24 18:38:17 UTC
assuming so