[Qemu-devel] Re: Re: Firefox crashes inside qemu

["Alexander E. Patrakov" <patrakov@ums.usu.ru>]

Alexander E. Patrakov wrote:

> Lennert Buytenhek wrote:
> 
>> On Sun, Jan 02, 2005 at 08:36:30PM +0500, Alexander E. Patrakov wrote:
>> 
>>> >> Firefox will crash. I suppose that there is some bug in the CPU
>>> >> emulation.
>>> >> 
>>> >> How can I help debugging this problem further?
>>> > 
>>> > Can you try that disk image on a 'real' machine and see if it crashes
>>> > then also?
>>> 
>>> I instaled win2k from the same CD to a real computer. Firefox downloaded
>>> from the same URL doesn't crash under the same condition.
>> 
>> There are many different factors in both scenarios besides that one of
>> them uses qemu and the other one does not: disk size, video card, network
>> card, etc.
>> 
>> 
>>> But how do I test a qemu image on a real machine?
>> 
>> I would just dd it to an old IDE disk, even though that might give issues
>> with differing head/sector counts.
>> 
>> dd if=disk.img of=/dev/hdx bs=1024k
>> 
>> (Where /dev/hdx is the old IDE disk.)
> 
> I see. The problem is that I don't have any old IDE disk at hand. But I
> will dd that into VMware tomorrow in order to see if the installation is
> really bad. As for the geometry, the doc at
> http://linux-ntfs.sourceforge.net/ntfs/files/boot.html is known sufficient
> for fixing such issues.
> 

After changing the IDE driver to "generic dual-channel" (in order to avoid
"inaccessible boot device" blue screen) and dd'ing the disk image into
VMware (using Knoppix and ssh to the host), Firefox doesn't crash. So the
problem is indeed in qemu.

I also had private mail conversation with marc, who suggested to try his
"interp" patch. Since qemu cannot boot win2k with this patch, I had to boot
win2k with the regular qemu, savevm, loadvm into patched qemu and only then
start Firefox. It then crashed as described in the first mail in this
thread.

I hope this information is useful.

-- 
Alexander E. Patrakov