Rosegarden

On Saturday 26 February 2005 20:52, Guillaume Laurent wrote:
> - from the top level dir, run ./scons_configure (you only need to do this
> once, of course)

I can't find "sconf_configure" anywhere. Is it supposed to be in RG CVS now?

Regards,
Pedro

On Saturday 26 Feb 2005 19:52, Guillaume Laurent wrote:
> - from the top level dir, run ./scons_configure

Did you commit that?


Chris

On Saturday 26 February 2005 21:38, Pedro Lopez-Cabanillas wrote:
> On Saturday 26 February 2005 20:52, Guillaume Laurent wrote:
> > - from the top level dir, run ./scons_configure (you only need to do this
> > once, of course)
>
> I can't find "sconf_configure" anywhere. Is it supposed to be in RG CVS
> now?

It's supposed to be, as this message from rg-bugs testifies, but to my 
amazement it is indeed not :

https://sourceforge.net/mailarchive/message.php?msg_id=11011845

I had the same kind of problems with another file while trying to commit this 
set of changes. Seems that sforge is having hiccups again, fairly serious 
ones.

-- 
      Guillaume.
      http://www.telegraph-road.org

On Saturday 26 Feb 2005 19:52, Guillaume Laurent wrote:
> It works now. All you need to do is :
>
> - install scons (http://scons.org)
> - from the top level dir, run ./scons_configure [...]
> - run scons

Not working for me:

- snip -

studio@...]$ ./scons_configure

This is bksys 1.1.0


Start 'scons' or 'make' now (and 'scons install' or 'make install' as root to 
install)
Make sure to read the output of ./configure --help if an error occurs
studio@...]$ scons
scons: Reading SConscript files ...
IndentationError: expected an indented block (config.py, line 9)

studio@...]$ cat config.py
#
# script generated by configure
# Sun Feb 27 09:58:50 2005
#

def exists():
        return True
def generate(env):
studio@...]$

- snip -

As someone who has always despised Python's indentation-as-syntax thing, I am 
of course delighted to see that the reported error is an indentation one.  
(Even though in practice it's presumably caused by a premature EOF.)

One real irritation with scons is that it won't do anything -- not even parse 
its options -- if its config.py is busted.  So in this case I can't even type 
"scons --help" or something to try to work out whether I've done something 
wrong, because it just barfs immediately with the same error about config.py.

btw, the invocation of scons_configure returned so quickly that I suspected it 
probably hadn't done anything even before I subsequently ran scons.  I see 
scons_configure is looking for bk*.py -- I have bksys.py, but nothing else 
that matches.


Chris

On Sunday 27 February 2005 11:08, Chris Cannam wrote:
> On Saturday 26 Feb 2005 19:52, Guillaume Laurent wrote:
> > It works now. All you need to do is :
> >
> > - install scons (http://scons.org)
> > - from the top level dir, run ./scons_configure [...]
> > - run scons
>
> Not working for me:

*sigh*. Sorry, there was a bkkde.py which I forgot to commit. It's in, should 
work now.

-- 
      Guillaume.
      http://www.telegraph-road.org

Hi,

scons_configure is there! I guess that you were adding the files to the root 
directory instead of the 'rosegarden' module?
http://cvs.sourceforge.net/viewcvs.py/rosegarden/

Anyway, the file 'bkkde.py' is missing. Without it, the file 'config.py' was 
wrong and scons didn't run. I added it from the bksys-1.1 distribution, and 
the configuration was smoothly writting 'config.py'.

Next step. Running scons :
$ scons
scons: Reading SConscript files ...
Checking for main() in C library alsa... no
Checking for main() in C library jack... yes
Checking for C header file ladspa.h... yes
Checking for main() in C library lo... yes
Checking for main() in C library rdf... no
Checking for main() in C library mad... yes
Checking for main() in C library dssi... no

scons: warning: Ignoring missing SConscript 'doc/en/SConscript'
File "SConstruct", line 66, in ?

scons: warning: Ignoring missing SConscript 'po/SConscript'
File "SConstruct", line 67, in ?
scons: done reading SConscript files.
scons: Building targets ...
g++ -DHAVE_ALSA -DQT_THREAD_SUPPORT -DHAVE_JACK -DHAVE_LADSPA -DHAVE_LIBLO 
-DHAVE_LIBMAD -I/usr/include/ -I/usr/lib/qt3//include/ -I. -Ibase -c -o 
base/AnalysisTypes.o base/AnalysisTypes.C

Aha, seems working, but I have ALSA, liblrdf and dssi installed, and scons 
only recognizes ALSA.

The compilation failed in gui/compositionview.cpp, but it seems a source 
problem.

gui/compositionview.cpp: In member function `virtual 
std::vector<CompositionItem, std::allocator<CompositionItem> > 
CompositionModelImpl::getItemsAt(const QPoint&)':
gui/compositionview.cpp:345: error: no match for 'operator<<' in 
'(&kdDebug(int)())->kdbgstream::operator<<(const 
char*)("CompositionModelImpl::getItemsAt() adding ") << sr'

Regards,
Pedro

Guillaume Laurent wrote:
>
> It works now. All you need to do is :
> 
> - install scons (http://scons.org)
> - from the top level dir, run ./scons_configure (you only need to do
> this once, of course)
> 
> - run scons

It doesn't work for me.

  ./scons_configure

Traceback (most recent call last):
  File "./scons_configure", line 12, in ?
    import bksys
ImportError: No module named bksys

This is despite having supposedly installed bksys - at least according to
its limited documentation.

William

On Sunday 27 February 2005 11:40, Pedro Lopez-Cabanillas wrote:
> Hi,
>
> scons_configure is there! I guess that you were adding the files to the
> root directory instead of the 'rosegarden' module?
> http://cvs.sourceforge.net/viewcvs.py/rosegarden/

Strange. I had temporarily disabled the 'update recursively' setting in 
cervisia before adding and committing those files the first time, may be that 
was the reason. That doesn't explain why SConstruct is here too, since I 
committed that one in no different way as the others a while ago... No idea.

> Anyway, the file 'bkkde.py' is missing. Without it, the file 'config.py'
> was wrong and scons didn't run. I added it from the bksys-1.1 distribution,
> and the configuration was smoothly writting 'config.py'.

Yes, just added it.

> Aha, seems working, but I have ALSA, liblrdf and dssi installed, and scons
> only recognizes ALSA.

Yes, lib detection isn't perfected yet (or I don't know how to use it 
properly)

> gui/compositionview.cpp: In member function `virtual
> std::vector<CompositionItem, std::allocator<CompositionItem> >
> CompositionModelImpl::getItemsAt(const QPoint&)':
> gui/compositionview.cpp:345: error: no match for 'operator<<' in
> '(&kdDebug(int)())->kdbgstream::operator<<(const
> char*)("CompositionModelImpl::getItemsAt() adding ") << sr'

Which version of gcc and KDE libs do you have installed ? This line compiles 
fine here. Anyway it's just a temporary trace so you can comment it out.

-- 
      Guillaume.
      http://www.telegraph-road.org

Guillaume Laurent wrote:
>
> *sigh*. Sorry, there was a bkkde.py which I forgot to commit. It's in,
> should work now.

It still doesn't work for me.  I forgot to mention in my previous message
that the error I get below is after having installed your bkkde.py from cvs.

William

On Sun, 27 Feb 2005 10:41:21 GMT, William wrote
> Guillaume Laurent wrote:
>>
>> It works now. All you need to do is :
>>
>> - install scons (http://scons.org)
>> - from the top level dir, run ./scons_configure (you only need to do
>> this once, of course)
>>
>> - run scons
>
> It doesn't work for me.
>
>  ./scons_configure
>
> Traceback (most recent call last):
>   File "./scons_configure", line 12, in ?
>     import bksys
> ImportError: No module named bksys
>
> This is despite having supposedly installed bksys - at least according to
> its limited documentation.

On Sunday 27 February 2005 11:54, Guillaume Laurent wrote:
> > Aha, seems working, but I have ALSA, liblrdf and dssi installed, and
> > scons only recognizes ALSA.
>
> Yes, lib detection isn't perfected yet (or I don't know how to use it
> properly)

I will try to play with it a little.

> > gui/compositionview.cpp: In member function `virtual
> > std::vector<CompositionItem, std::allocator<CompositionItem> >
> > CompositionModelImpl::getItemsAt(const QPoint&)':
> > gui/compositionview.cpp:345: error: no match for 'operator<<' in
> > '(&kdDebug(int)())->kdbgstream::operator<<(const
> > char*)("CompositionModelImpl::getItemsAt() adding ") << sr'
>
> Which version of gcc and KDE libs do you have installed ? This line
> compiles fine here. Anyway it's just a temporary trace so you can comment
> it out.

Qt: 3.1.2
KDE: 3.1.3
gcc version 3.3.1 (Mandrake Linux 9.2 3.3.1-2mdk)

BTW, scons_configure screams (in red) that my KDE 3.1 can be "too old". Sigh.

Regards,
Pedro

On Sunday 27 Feb 2005 11:03, William wrote:
> Guillaume Laurent wrote:
> > *sigh*. Sorry, there was a bkkde.py which I forgot to commit. It's
> > in, should work now.
>
> It still doesn't work for me.  I forgot to mention in my previous
> message that the error I get below is after having installed your
> bkkde.py from cvs.

bksys.py should be in the top-level RG directory now -- Guillaume added 
it yesterday or today or whenever.  Check you have it.

The previous install of bksys that you made is (AFAICS) now irrelevant.


Chris

On Sunday 27 Feb 2005 10:54, Guillaume Laurent wrote:
> > Aha, seems working, but I have ALSA, liblrdf and dssi installed,
> > and scons only recognizes ALSA.
>
> Yes, lib detection isn't perfected yet (or I don't know how to use it
> properly)

Hm, well it is obviously kind of fundamental that we should be able to 
easily add in dependencies for other libraries.  That's about the only 
thing I find easy to do with autoconf.

In this case, liblrdf and DSSI both use pkg-config, so they should be as 
easy to handle as anything new could possibly be.

(btw our current configure.in.in includes an extra test for versions of 
liblo beyond a certain minor number -- 0.7 or something.  I don't mind 
throwing away that test, I think older versions are probably already 
scarce.)


Chris

Chris Cannam wrote:
>
> bksys.py should be in the top-level RG directory now -- Guillaume added 
> it yesterday or today or whenever.  Check you have it.

Thanks for the tip; I had bksys.py in local CVS but not in my "most local" CVS.
Unfortunately despite having it now, the scons build still fails:

 $ scons
scons: Reading SConscript files ...

scons: *** No tool named 'kde': No module named kde
File "SConstruct", line 11, in ?

William

On Sunday 27 February 2005 12:10, Pedro Lopez-Cabanillas wrote:
> 
> BTW, scons_configure screams (in red) that my KDE 3.1 can be "too old".

Er, 3.4 is just about to be released, and Qt's current version is 3.3.x. :-)

-- 
      Guillaume.
      http://www.telegraph-road.org

On Sunday 27 February 2005 12:07, Chris Cannam wrote:
> 
> The previous install of bksys that you made is (AFAICS) now irrelevant.

There's nothing to install with bksys actually, its components are meant to be 
copied in your own project. Just download it for reference, but all the files 
you should need are in RG's top level dir (plus scons itself, of course).

-- 
      Guillaume.
      http://www.telegraph-road.org

On Sunday 27 February 2005 12:18, William wrote:
> Chris Cannam wrote:
> > bksys.py should be in the top-level RG directory now -- Guillaume added
> > it yesterday or today or whenever.  Check you have it.
>
> Thanks for the tip; I had bksys.py in local CVS but not in my "most local"
> CVS. Unfortunately despite having it now, the scons build still fails:

Check that you have bkkde.py too.

-- 
      Guillaume.
      http://www.telegraph-road.org

On Sunday 27 Feb 2005 11:28, Guillaume Laurent wrote:
> On Sunday 27 February 2005 12:10, Pedro Lopez-Cabanillas wrote:
> > BTW, scons_configure screams (in red) that my KDE 3.1 can be "too
> > old".
>
> Er, 3.4 is just about to be released, and Qt's current version is
> 3.3.x. :-)

Yeah, but if it's new enough for RG, it's got to be new enough for RG's 
build system!


Chris

Guillaume Laurent wrote:
> On Sunday 27 February 2005 12:18, William wrote:
>> Chris Cannam wrote:
>> > bksys.py should be in the top-level RG directory now -- Guillaume added
>> > it yesterday or today or whenever.  Check you have it.
>>
>> Thanks for the tip; I had bksys.py in local CVS but not in my "most local"
>> CVS. Unfortunately despite having it now, the scons build still fails:
> 
> Check that you have bkkde.py too.

I do. It's from your version with modification time 10:34:55 today.

William

On Sunday 27 February 2005 12:09, Chris Cannam wrote:
> On Sunday 27 Feb 2005 10:54, Guillaume Laurent wrote:
> > > Aha, seems working, but I have ALSA, liblrdf and dssi installed,
> > > and scons only recognizes ALSA.
> >
> > Yes, lib detection isn't perfected yet (or I don't know how to use it
> > properly)
>
> Hm, well it is obviously kind of fundamental that we should be able to
> easily add in dependencies for other libraries.  That's about the only
> thing I find easy to do with autoconf.

Patch attached. It tests for a minimal ALSA library version, and corrects 
another few tests.

Regards,
Pedro

On Sunday 27 Feb 2005 11:48, Pedro Lopez-Cabanillas wrote:
> Patch attached. It tests for a minimal ALSA library version, and
> corrects another few tests.

I still don't trust this.  It should be using pkg-config in the first 
instance for libraries that use it, and then checking the headers or 
libraries using the include or link paths returned by pkg-config.

Maybe it is doing that -- I haven't looked at the code.  I'm just wary 
of the fact that it only prints "Checking for main() in C library blah" 
and nothing else.

btw, we still also appear to lack a test for Xft.


Chris

On Sunday 27 February 2005 12:50, Chris Cannam wrote:
> On Sunday 27 Feb 2005 11:48, Pedro Lopez-Cabanillas wrote:
> > Patch attached. It tests for a minimal ALSA library version, and
> > corrects another few tests.
>
> I still don't trust this.  It should be using pkg-config in the first
> instance for libraries that use it, and then checking the headers or
> libraries using the include or link paths returned by pkg-config.

Trying to find references to solve your complaints, I've found that 
cheesetracker is using Scons as a build system. This project uses Qt, ALSA, 
Jack, LADSPA and tests dependencies using pkg-config.  

I will try to look it deeply.
http://www.reduz.com.ar/cheesetronic/

Regards,
Pedro

On Sunday 27 February 2005 13:35, I wrote:
> Trying to find references to solve your complaints, I've found that
> cheesetracker is using Scons as a build system. This project uses Qt, ALSA,
> Jack, LADSPA and tests dependencies using pkg-config.
>
> I will try to look it deeply.
> http://www.reduz.com.ar/cheesetronic/

An this remembers me that Juan Linietsky was a subscriber on this list not a 
long time. 

¿Que tal, Juan? ;-)

Saludos,
Pedro

On Sunday 27 February 2005 12:44, William wrote:
> Guillaume Laurent wrote:
> 
> > Check that you have bkkde.py too.
>
> I do. It's from your version with modification time 10:34:55 today.

Have you re-run scons_configure ? I double-checked and everything is in.

-- 
      Guillaume.
      http://www.telegraph-road.org

On Sunday 27 February 2005 12:48, Pedro Lopez-Cabanillas wrote:
>
> Patch attached. It tests for a minimal ALSA library version, and corrects
> another few tests.

Thanks, I applied it. One thing which is lacking is that there doesn't seem to 
be a way how to tell scons to look for a lib in specific dirs, when the lib 
doesn't use pkg-config, like Xft for instance.

-- 
      Guillaume.
      http://www.telegraph-road.org

Guillaume Laurent wrote:
>
> On Sunday 27 February 2005 12:44, William wrote:
>> Guillaume Laurent wrote:
>> 
>> > Check that you have bkkde.py too.
>>
>> I do. It's from your version with modification time 10:34:55 today.
> 
> Have you re-run scons_configure ? I double-checked and everything is in.

It turned out "kde.py" was missing in my local CVS.
The build still fails though exactly as it did on 21st February in notefont.cpp
because -I/usr/X11R6/include is still missing from the include paths:

http://sourceforge.net/mailarchive/message.php?msg_id=10924184

g++ -DHAVE_ALSA -DQT_THREAD_SUPPORT -DHAVE_JACK -DHAVE_LADSPA -DHAVE_LIBLO -DHAVE_LIBMAD -I/rem/kde/include/ -I/rem/lib/qt/include/ -I/usr/X11R6/include/X11/Xft -I/usr/include/freetype2 -Ibase -Isound -I. -Igui -c -o gui/notefont.o gui/notefont.cpp
In file included from gui/notefont.cpp:59:
/usr/include/X11/Xft/Xft.h:42:35: fontconfig/fontconfig.h: No such file or directory

I notice you have some hard-coded include paths in gui/SConstrict 
and if I manually add /usr/X11R6/include, the build gets further
before it fails again due to missing "config.h" includes.

William

On Sunday 27 February 2005 16:43, Guillaume Laurent wrote:
> Thanks, I applied it. One thing which is lacking is that there doesn't seem
> to be a way how to tell scons to look for a lib in specific dirs, when the
> lib doesn't use pkg-config, like Xft for instance.

Xft uses pkg-config here:

$ pkg-config xft --libs
-L/usr/X11R6/lib -lXft -lX11 -lfreetype -lXrender -lfontconfig
$ pkg-config xft --modversion
2.1.2

Regards,
Pedro

On Sunday 27 February 2005 16:46, William wrote:
> 
> It turned out "kde.py" was missing in my local CVS.

It should be generated by scons_configure.

> The build still fails though exactly as it did on 21st February in
> notefont.cpp because -I/usr/X11R6/include is still missing from the include
> paths:

OK, I'll investigate later, right now I want to finish fixing the repeat 
segment pbs.

-- 
      Guillaume.
      http://www.telegraph-road.org

On Sunday 27 February 2005 16:54, Pedro Lopez-Cabanillas wrote:
>
> Xft uses pkg-config here:

Duh, here too. When I tried getting it to detect Xft and it just wouldn't find 
it I thought it was just some silly path problem.

-- 
      Guillaume.
      http://www.telegraph-road.org

Guillaume Laurent wrote:
>
> On Sunday 27 February 2005 16:46, William wrote:
>> 
>> It turned out "kde.py" was missing in my local CVS.
> 
> It should be generated by scons_configure.

It doesn't seem to be for me, so I created one by copying yours from CVS.

>> The build still fails though exactly as it did on 21st February in
>> notefont.cpp because -I/usr/X11R6/include is still missing from the include
>> paths:
> 
> OK, I'll investigate later, right now I want to finish fixing the repeat 
> segment pbs.

Ok.

William

On Sunday 27 February 2005 17:23, William wrote:
> Guillaume Laurent wrote:
> > On Sunday 27 February 2005 16:46, William wrote:
> >> It turned out "kde.py" was missing in my local CVS.
> >
> > It should be generated by scons_configure.
>
> It doesn't seem to be for me, so I created one by copying yours from CVS.

Woops, you're right, I was misled by the comment at the head of kde.py. I'll 
report that to Thomas.

-- 
      Guillaume.
      http://www.telegraph-road.org

On Sunday 27 February 2005 06:31 am, Guillaume Laurent wrote:
> On Sunday 27 February 2005 12:07, Chris Cannam wrote:
> > The previous install of bksys that you made is (AFAICS) now irrelevant.
>
> There's nothing to install with bksys actually, its components are meant to
> be copied in your own project. Just download it for reference, but all the
> files you should need are in RG's top level dir (plus scons itself, of
> course).

I'm only half following this, and can't be bothered to dig back and see how 
you said to fix this.  I:

make distclean
./scons_configure
scons

The configure step worked, I guess.  Purty colors.  Make puked on 
base/config.h which IIRC you said to fix by running a regular automake based 
configure, but I also thought you said you had hacked around that necessity.

If it matters:

->dpkg -l|grep scons
ii  scons          0.96.1-1       A replacement for Make


ar: creating base/libbase.a
g++ -DHAVE_ALSA -DQT_THREAD_SUPPORT -DHAVE_JACK -DHAVE_LADSPA -DHAVE_LIBLO 
-DHAVE_LIBMAD -I/usr/include/kde/ -I/usr/include/qt3 
-I/usr/X11R6/include/X11/Xft -I/usr/include/freetype2 -Ibase -Isound -I. 
-Igui -c -o gui/audiocommands.o gui/audiocommands.cpp
/usr/bin/moc -o gui/audiomanagerdialog.moc gui/audiomanagerdialog.h
g++ -DHAVE_ALSA -DQT_THREAD_SUPPORT -DHAVE_JACK -DHAVE_LADSPA -DHAVE_LIBLO 
-DHAVE_LIBMAD -I/usr/include/kde/ -I/usr/include/qt3 
-I/usr/X11R6/include/X11/Xft -I/usr/include/freetype2 -Ibase -Isound -I. 
-Igui -c -o gui/audiomanagerdialog.o gui/audiomanagerdialog.cpp
In file included from gui/rosegardenguidoc.h:36,
                 from gui/audiomanagerdialog.h:28,
                 from gui/audiomanagerdialog.cpp:44:
gui/rosegardengui.h:25:20: config.h: No such file or directory
gui/audiomanagerdialog.cpp: In member function `void
   Rosegarden::AudioManagerDialog::slotRename()':

-- 
Michael McIntyre  ----   Silvan <dmmcintyr@...>
Linux fanatic, and certified Geek;  registered Linux user #243621
http://www.geocities.com/Paris/Rue/5407/
http://rosegarden.sourceforge.net/tutorial/

On Sunday 27 February 2005 17:52, Silvan wrote:
>
> The configure step worked, I guess.  Purty colors.  Make puked on
> base/config.h which IIRC you said to fix by running a regular automake
> based configure, but I also thought you said you had hacked around that
> necessity.

No, I haven't yet. Just dug up a config.h generation method from the scons 
wiki, I'll use it. So for the moment we're still depending on a generated 
config.h

-- 
      Guillaume.
      http://www.telegraph-road.org

On Sunday 27 February 2005 12:04 pm, Guillaume Laurent wrote:
> On Sunday 27 February 2005 17:52, Silvan wrote:
> > The configure step worked, I guess.  Purty colors.  Make puked on
> > base/config.h which IIRC you said to fix by running a regular automake
> > based configure, but I also thought you said you had hacked around that
> > necessity.
>
> No, I haven't yet. Just dug up a config.h generation method from the scons
> wiki, I'll use it. So for the moment we're still depending on a generated
> config.h

OK, I ./configured it, then tried scons again.  Now:

gui/rosegardenconfiguredialog.cpp:2628:43: rosegardenconfigurationpage.moc: No 
such file or directory
scons: *** [gui/rosegardenconfiguredialog.o] Error 1
scons: building terminated because of errors.

-- 
Michael McIntyre  ----   Silvan <dmmcintyr@...>
Linux fanatic, and certified Geek;  registered Linux user #243621
http://www.geocities.com/Paris/Rue/5407/
http://rosegarden.sourceforge.net/tutorial/

On Sunday 27 Feb 2005 17:24, Silvan wrote:
> gui/rosegardenconfiguredialog.cpp:2628:43:
> rosegardenconfigurationpage.moc: No such file or directory

Yeah, I got that too -- didn't report it or get time to look closer yet.


Chris

On Sunday 27 February 2005 19:12, Chris Cannam wrote:
> On Sunday 27 Feb 2005 17:24, Silvan wrote:
> > gui/rosegardenconfiguredialog.cpp:2628:43:
> > rosegardenconfigurationpage.moc: No such file or directory
>
> Yeah, I got that too -- didn't report it or get time to look closer yet.

I see what the problem is, fix coming.

-- 
      Guillaume.
      http://www.telegraph-road.org

On Sunday 27 February 2005 12:50, Chris Cannam wrote:
> I still don't trust this.  It should be using pkg-config in the first
> instance for libraries that use it, and then checking the headers or
> libraries using the include or link paths returned by pkg-config.

OK. Committed an initial test set using pkg-config.

The libraries and versions tested with pkg-config are:
	alsa	1.0 
	jack	0.77
	liblo	0.7
	mad	0.10
	dssi	0.4
	xft	2.1.0

Regards,
Pedro

On Sunday 27 February 2005 19:21, Guillaume Laurent wrote:
> On Sunday 27 February 2005 19:12, Chris Cannam wrote:
> > On Sunday 27 Feb 2005 17:24, Silvan wrote:
> > > gui/rosegardenconfiguredialog.cpp:2628:43:
> > > rosegardenconfigurationpage.moc: No such file or directory
> >
> > Yeah, I got that too -- didn't report it or get time to look closer yet.
>
> I see what the problem is, fix coming.

Fix is in. I just moved all the configuration page classes to their own file, 
which we should have done long ago.

-- 
      Guillaume.
      http://www.telegraph-road.org