Bug#535858: iceweasel forgets first page setting on quit

[Mike Hommey]

On Wed, Dec 23, 2009 at 12:21:10PM +0100, Peter T. Breuer wrote:
> "Also sprach Mike Hommey:"
> > Sorry for the late answer.
> 
> No problem ...
> 
> > On Sun, Jul 05, 2009 at 06:22:21PM +0200, Peter T. Breuer wrote:
> > > Package: iceweasel
> > > Version: 3.0.11-1
> > > Severity: normal
> > > 
> > > 
> > > After quiting firefox (usually using the window X button, not quit from
> > > the menu ...) firefox pretty well always reverts next time to showing
> > > my "windows and tabs from last time" at startup, though I always set
> > > the preference in edit preferences to "show blank page" (as appropriate
> > > for a machine that isn't always on the net). It's annoying!
> > > 
> > > This started happening within the last month, after a recent weasel
> > > upgrade, or maybe after the upgrade which took me to libc 2.9 (and
> > > which borke cut and paste from an Eterm).
> > 
> > This sounds like your iceweasel instance is crashing when you kill it.
> 
> Well, it doesn't look that way. It usually asks if I want to close all
> those tabs at once ...

It can still crash after that. That did happen some time ago, where some
crash could happen after libxul.so is dlclose()d, for example.

> > First, does it still happen for you ? If it does, could you try to
> > attach a debugger before killing iceweasel ?
> 
> OK, I can do a lot of debugging. I'll let you know. I guess you're not
> looking forward to closing your iceweasel down either!
> 
> > Cheers,
> > 
> > Mike
> > 
> > PS: I see you are an Ubuntu user, did you really install the iceweasel
> > package?
> 
> I'm not an ubuntu user. That distro name stuff seems to come from
> some place that I've never bothered to discover. Cat
> /ect/debian_version says
> 
>   squeeze/sid

I'm asking because your bug report said:

Distributor ID: Ubuntu
Description:    Ubuntu 8.04.1
Release:        8.04
Codename:       hardyArchitecture: i386 (i686)


> 
> Which also can't be quite right. I follow debian unstable. dpkg -l
> iceweasel says:
> 
>  ii  iceweasel      3.5.6-1        lightweight web browser based on Mozilla
> 
> and "iceweasel" is how I invoke it. Looking at the link:
> 
>   % ll `which iceweasel`
>   lrwxrwxrwx 1 root root 26 Dec 22 23:04 /usr/bin/iceweasel -> ../lib/iceweasel/iceweasel*
> 
> Which looks conclusive. However, the executable when in flight calls
> itself "firefox-bin" in the process table:
> 
>                   |-firefox-bin-+-3*[kpdf]
>         |         |             `-14*[{firefox-bin}]
> 
>  
> and that seems to be because "iceweasel" is in fact a shell script:
> 
>   #!/bin/sh
>   #
>   # The contents of this file are subject to the Netscape Public
>   # License Version 1.1 ...
> 
> and it looks as though it launches:
> 
>  MOZ_DIST_BIN="/usr/lib/iceweasel"
>  MOZ_PROGRAM="${MOZ_DIST_BIN}/firefox-bin"    <---- that
>  MOZ_APP_LAUNCHER=/usr/bin/iceweasel
> 
> So /usr/lib/iceweasel/firefox-bin is what's in the air. Dpkg -S says:
> 
>  iceweasel: /usr/lib/iceweasel/firefox-bin
> 
> However, that's merely ANOTHER link:
> 
>  lrwxrwxrwx 1 root root 33 Dec 22 23:04 /usr/lib/iceweasel/firefox-bin -> ../xulrunner-1.9.1/xulrunner-stub*
> 
> And dpkg says:
> 
>  ii  xulrunner-1.9. 1.9.1.6-1      XUL + XPCOM application runner
> 
> Whatever an xul is, I don't know. But xulrunner is also a debian
> package. It's not big on man pages or docs, but the docs I can see
> point squarely at a debian origin. Mozilla public license. Note
> from some stranger called "Mike Hommey" at debian in 2006 as the
> README.Debian.  Debian changelog updated by yourself in 16 Dec.
> 
> As to what it actually IS, not a clue available (situation normal:
> grumble grumble, a README used to actually SAY).
> 
> BTW, the iceweasel script seems to be dedicated to intercepting a
> --debugger flag and launching the program with gdb or ddd as
> controller. One might have thought a good attempt at that could be
> made from within firefox itself, launching a debugger window
> containing a ptrace attached to itself (assuming it gets that far,
> that is). It would have been user-friendlier, in a sense.

That's what bug-buddy and other programs are for.

Cheers,

Mike