[Qemu-devel] Qemu frontend proposal

[Qemu-devel] Qemu frontend proposal

[Jean-Michel POURE]

Dear all,

Fabrice expressed the need for a Qemu frontend. Before working on anything, I 
would like to discuss about possible plans and receive some feedback.

My proposals are:

1) Framework : wxWidgets

wxWidgets is the new name of the wxWindows framework. The latest release 
wxWidgets 2.5.1 has good Unicode compliance and offers GTK2 and Windows 
bindings. The wxMac version is still hardly usable.

Also, wxWidgets offers a wxPython binding.

2) Interface builder : wxGlade

wxWidgets supports an XML format called XRC, which is called at runtime 
without the need to hard-code the interface.

There are several applications for building XRC interface. wxGlade is probably 
the best and very similar to Glade-2.

Under Debian, I installed wxPython 2.5.1 from the experimental repository and 
ran wxGlade from CVS version without real problem.

3) Collaboration : group work inside the Qemu project

I would be in favour of working in team inside the Qemy project. My guess 
would be to discuss the interface first, commit the XRC to Qemu CVS. Then, we 
can start coding the underlying logic in C++ or in Python.

This way, we can share ideas and experience. Any feedback is appreciated. Do 
not hesitate to answer on the list.

Cheers,
Jean-Michel

Re: [Qemu-devel] Qemu frontend proposal

[Kyle Hayes]

On Friday 23 April 2004 00:13, Jean-Michel POURE wrote:
> Dear all,
>
> Fabrice expressed the need for a Qemu frontend. Before working on
> anything, I would like to discuss about possible plans and receive
> some feedback.
>
> My proposals are:
>
> 1) Framework : wxWidgets
> 2) Interface builder : wxGlade

There are two perfectly good toolkits that are on nearly every distro 
out of the box right now:  GTK and Qt.  Both have excellent tools and 
language bindings and interface builders.  Python has perfectly good 
bindings for Qt and GTK on its own.  I see no functional need at all 
to introduce yet another toolkit.

Building open source projects is already getting to be more and more 
annoying as each one wants to bring in another fifty dependencies 
that are not usually installed.  And, each one wants to wrap some 
perfectly good libraries in their own just-barely-different wrappers.  
The more dependencies everything has the more likely we'll all end up 
in the Linux version of Windows DLL hell.

Please, please, please think twice about introducing new dependencies!

What are the goals for a QEMU interface?  Here's what I'd like to see:

1) portable.

2) very lightweight.

3) remotely accessible.

4) easily maintained and changed.

5) i18n.

What this says to me is that a web interface would cover what I want.  
And, that is guaranteed to be portable.   In my experience, an 
embedded web interface tends to be smaller (code size) than any 
toolkit-based code.    When the browser isn't up, you aren't using 
that memory.

Better yet, make the interface in QEMU in XML-RPC and then everyone 
and their dog can write an interface in any language or system.  
Please stay with the Unix philosophy of making each tool do one thing 
well.

Best,
Kyle

Re: [Qemu-devel] Qemu frontend proposal

[Michael L Torrie]

On Fri, 2004-04-23 at 01:13, Jean-Michel POURE wrote:
> Dear all,
> 
> Fabrice expressed the need for a Qemu frontend. Before working on anything, I 
> would like to discuss about possible plans and receive some feedback.

As good as wxWidgets and other GUI toolkits are, one size does not fit
all.  If you are trying to reach win32, *nix, and OS X, right now only
QT comes close but there is no free version currently available on win32
(well, there is a GPL port of the X11 version to win32 and it's quite
good, but still under heavy development).  GTK 2 looks good on windows
and unix now, but it's a no show under OS X.  The pros and cons of
wxWindows you have already dealt with.

So, from watching various programs that are trying to implement
universal front-ends for programs, I've found that often the best
solution (especially for programs with simpler UI needs like qemu) is to
have a clean separation between the backend and frontend.

With that in mind, a wxWidgets gui is a great idea, so long as it's easy
and clean for someone (if they desired) to create a new GUI with some
other toolkit and interface cleanly with the qemu backend.  In other
words, if all the backend needs is a SDL window handle and then an API
for communicating with the GUI, that is the best, in my opinion.  The
GUI itself should not be an integral part of the qemu backend. 
Currently there is an API of sorts for use by the front end by way of
standard-in, standard-out.  Plugging into this currently is true to the
unix spirit.

Isn't there a win32 frontend currently, and work in a unix port of it?

Michael

> 
> My proposals are:
> 
> 1) Framework : wxWidgets
> 
> wxWidgets is the new name of the wxWindows framework. The latest release 
> wxWidgets 2.5.1 has good Unicode compliance and offers GTK2 and Windows 
> bindings. The wxMac version is still hardly usable.
> 
> Also, wxWidgets offers a wxPython binding.
> 
> 2) Interface builder : wxGlade
> 
> wxWidgets supports an XML format called XRC, which is called at runtime 
> without the need to hard-code the interface.
> 
> There are several applications for building XRC interface. wxGlade is probably 
> the best and very similar to Glade-2.
> 
> Under Debian, I installed wxPython 2.5.1 from the experimental repository and 
> ran wxGlade from CVS version without real problem.
> 
> 3) Collaboration : group work inside the Qemu project
> 
> I would be in favour of working in team inside the Qemy project. My guess 
> would be to discuss the interface first, commit the XRC to Qemu CVS. Then, we 
> can start coding the underlying logic in C++ or in Python.
> 
> This way, we can share ideas and experience. Any feedback is appreciated. Do 
> not hesitate to answer on the list.
> 
> Cheers,
> Jean-Michel
> 
> 
> _______________________________________________
> Qemu-devel mailing list
> Qemu-devel@nongnu.org
> http://mail.nongnu.org/mailman/listinfo/qemu-devel
-- 
Michael L Torrie <torriem@chem.byu.edu>

Re: [Qemu-devel] Qemu frontend proposal

[Jocelyn Mayer]

On Fri, 2004-04-23 at 09:13, Jean-Michel POURE wrote:
> Dear all,
> 
> Fabrice expressed the need for a Qemu frontend. Before working on anything, I 
> would like to discuss about possible plans and receive some feedback.
> 
[...]

I don't know much about existing frameworks, just a few things about
Gtk. I just want to point one requirement: I want and even sometime need
(!) that the choosen framework doesn't need X and can be used on the
Linux frame-buffer.

And, of course, it has to be easily portable and as light as possible...

Regards.

-- 
Jocelyn Mayer <l_indien@magic.fr>
Never organized

Re: [Qemu-devel] Qemu frontend proposal

[Fabrice Bellard]

Jean-Michel POURE wrote:
> Dear all,
> 
> Fabrice expressed the need for a Qemu frontend. Before working on anything, I 
> would like to discuss about possible plans and receive some feedback.
> 
> My proposals are:
 > [...]

Here is proposal:

1) Have an SDL interface which works for all OSes (Linux, Win32, MacOS 
X). SDL is useful to have a simple implementation to test a new port.

2) Make a GTK interface for Linux. It must be written in plain C and 
integrated in the QEMU CVS (a single C source file containing the UI 
might suffice for the first version). The default Linux build will be 
this one. I don't see the need of an external frontend or something 
written in another language.

3) Make a native win32 interface for Windows. It must compile with 
MinGW, so only the plain win32 API must be used.

Of course, in the cases (2) and (3), SDL won't be used. I will work on 
(2) and (3) if no one does it before me :-)

Fabrice.

Re: [Qemu-devel] Qemu frontend proposal

[Jean-Michel POURE]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

> Of course, in the cases (2) and (3), SDL won't be used. I will work on
> (2) and (3) if no one does it before me :-)

This is great, thanks!
Cheers, Jean-Michel
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)

iD8DBQFAihWKextoHHj2YFMRApuFAJ9Lrm+voxb1EawKk/A2d2c7zxLZXwCfWpJk
4GI3tNHsq4iJXKJGaFHiFFM=
=xYV+
-----END PGP SIGNATURE-----

Re: [Qemu-devel] Qemu frontend proposal

[Michael Torrie]

On Fri, 2004-04-23 at 16:18, Fabrice Bellard wrote:
> Here is proposal:
> 
> 1) Have an SDL interface which works for all OSes (Linux, Win32, MacOS 
> X). SDL is useful to have a simple implementation to test a new port.
> 
> 2) Make a GTK interface for Linux. It must be written in plain C and 
> integrated in the QEMU CVS (a single C source file containing the UI 
> might suffice for the first version). The default Linux build will be 
> this one. I don't see the need of an external frontend or something 
> written in another language.

There is a SDL widget for GTK.  Are you planning to use that at all or
are you going to re-implement the display (video) using pure GTK/GDK?

Michael

> 
> 3) Make a native win32 interface for Windows. It must compile with 
> MinGW, so only the plain win32 API must be used.
> 
> Of course, in the cases (2) and (3), SDL won't be used. I will work on 
> (2) and (3) if no one does it before me :-)
> 
> Fabrice.
> 
> 
> 
> _______________________________________________
> Qemu-devel mailing list
> Qemu-devel@nongnu.org
> http://mail.nongnu.org/mailman/listinfo/qemu-devel
-- 
Michael Torrie <torriem@chem.byu.edu>

Re: [Qemu-devel] Qemu frontend proposal

[Fabrice Bellard]

Michael Torrie wrote:
> On Fri, 2004-04-23 at 16:18, Fabrice Bellard wrote:
> 
>>Here is proposal:
>>
>>1) Have an SDL interface which works for all OSes (Linux, Win32, MacOS 
>>X). SDL is useful to have a simple implementation to test a new port.
>>
>>2) Make a GTK interface for Linux. It must be written in plain C and 
>>integrated in the QEMU CVS (a single C source file containing the UI 
>>might suffice for the first version). The default Linux build will be 
>>this one. I don't see the need of an external frontend or something 
>>written in another language.
> 
> 
> There is a SDL widget for GTK.  Are you planning to use that at all or
> are you going to re-implement the display (video) using pure GTK/GDK?

Using an SDL widget seems a bit overkill to me. I would prefer to 
reimplement the display in pure GTK/GDK.

Fabrice.

Re: [Qemu-devel] Qemu frontend proposal

[Martin Garton]

On Sun, 25 Apr 2004, Fabrice Bellard wrote:

> > There is a SDL widget for GTK.  Are you planning to use that at all or
> > are you going to re-implement the display (video) using pure GTK/GDK?
> 
> Using an SDL widget seems a bit overkill to me. I would prefer to 
> reimplement the display in pure GTK/GDK.

That would keep it more easily portable to other gtk+ platforms. AFAIK, 
gtksdl only works with x11 right now.  gtksdl could always be added later 
if pure gdk didn't perform.

-- 
Martin.