[Qemu-devel] vnc regression introduced by git commit 6baeb / svn r6860

[Qemu-devel] vnc regression introduced by git commit 6baeb / svn r6860

[Daniel Gollub]

Hi,

there seems to be a VNC regression introduced by git commit 6baeb (or svn 
r6860) when using vncviewer client (tightvnc). This seems to not affect all 
VNC clients, at least vinagre still works fine with latest QEMU trunk/master.


Already tried various version of tightvnc (1.3.9 and 1.3.10) - issues are the 
same. First vncviewer displays everything just fine - when it comes to a 
framebuffer update with a resolution resize it fails with:

$ vncviewer :1
[...]
Zero size rect - ignoring
Zero size rect - ignoring
Rect too large: 3156x11520 at (3156, 3328)
$

Result of bisecting this issue:

----->8------
commit 6baebed7698a37a0ac5168faf26023426b0ac940
Author: aliguori <aliguori@c046a42c-6fe2-441c-8c8c-71466251a162>
Date:   Fri Mar 20 15:59:14 2009 +0000

    vnc: cleanup surface handling, fix screen corruption bug. (Gerd Hoffmann)
-----8<----

Already tried to debug this - so far unfortunately without success.

It looks like the message which get retrieved by the vncviewer got malformed 
some how. The first rect is parsed fine by the vncviewer client, any further 
rect seems to malformed.

The source of the malformed message/package seems to be in vnc_update_client() 
when preparing and sending the screen update.

Not quite sure yet if this is a bug of tightvnc/vncviewer or QEMU, but since 
it's a regression i thought best thing is to do report it.

Best Regards,
Daniel

-- 
Daniel Gollub                        Geschaeftsfuehrer: Ralph Dehner
FOSS Developer                       Unternehmenssitz:  Vohburg
B1 Systems GmbH                      Amtsgericht:       Ingolstadt
Mobil: +49-(0)-160 47 73 970         Handelsregister:   HRB 3537
EMail: gollub@b1-systems.de          http://www.b1-systems.de

Adresse: B1 Systems GmbH, Osterfeldstraße 7, 85088 Vohburg
http://pgpkeys.pca.dfn.de/pks/lookup?op=get&search=0xED14B95C2F8CA78D

[Qemu-devel] Re: vnc regression introduced by git commit 6baeb / svn r6860

[Gerd Hoffmann]

   Hi,

> Rect too large: 3156x11520 at (3156, 3328)

Just noticed that too.  Wireshark reports strange rectangles as well. 
For some reason they are different from the ones vncviewer complains 
about through.  Investigating ...

cheers,
   Gerd

Re: [Qemu-devel] Re: vnc regression introduced by git commit 6baeb / svn r6860

[Anthony Liguori]

Gerd Hoffmann wrote:
>   Hi,
>
>> Rect too large: 3156x11520 at (3156, 3328)
>
> Just noticed that too.  Wireshark reports strange rectangles as well. 
> For some reason they are different from the ones vncviewer complains 
> about through.  Investigating ...
Are you using AutoSelect=0 FullColor=1?

Regards,

Anthony Liguori

> cheers,
>   Gerd
>
>

Re: [Qemu-devel] Re: vnc regression introduced by git commit 6baeb / svn r6860

[Yaniv Kaul]

On 4/27/2009 3:05 PM, Gerd Hoffmann wrote:
>   Hi,
>
>> Rect too large: 3156x11520 at (3156, 3328)
>
> Just noticed that too.  Wireshark reports strange rectangles as well. 
> For some reason they are different from the ones vncviewer complains 
> about through.  Investigating ...
And I thought it was a Wireshark bug. I'm still not convinced it isn't 
(The VNC dissector state machine isn't the greatest, and it lacks PDU 
reconstruction).
I also tried to push a patch to include dissection of VNC on 
non-standard ports 
(https://bugs.wireshark.org/bugzilla/show_bug.cgi?id=3215 ), which helps 
when connecting to virtual machines.
>
> cheers,
>   Gerd
>
>