On Tue, Oct 05, 2010 at 03:36:56PM +0200, Morten Kjærgaard wrote:
> Date: Tue, 5 Oct 2010 15:36:56 +0200
> From: Morten Kjærgaard <Morten.Kjargaard@prevas.dk>
> To: User discussions <ros-users@code.ros.org>
> Subject: Re: [ros-users] log4cxx dependency
> 
> >Why do you think log4cxx is unmaintained?  It hasn't had a release for a
> >while, but it's quite stable in our experience.  log4cxx was the best option
> >when first creating rosconsole ~2 years ago, and I haven't seen any new
> >software that changes that (boost.log has potential, but is unproven and
> >won't be available on most of our target platforms for years).
> 
> >Josh
> 
> Hi Josh
> 
> Normally I would expect software to be updated if it doesn't compile
> properly with new gcc versions. At least in my case (gcc 4.4.3) it
> required a little patching. I know it is easy to fix but it does
> give a signal of not being maintained.
> 
> I understand your argument about using a library with available
> packages for the different platforms, and that boost log has some
> drawbacks here. I am very interested in hearing about other
> possibilities.
> 
> My focus is more on cross compiling since I will be using ROS in
> commercial robot products. And I don't expect ROS to be build
> natively in these products. And in this case I found that log4cxx
> adds some unnecessary complexity, for adding such a "general" thing
> as logging.
> 

I could see hiding the dependency behind a buildtime switch, provided
this boost logging package can be wrapped to provide the same
interface (I haven't looked).  This switching can also be useful for
e.g. compiling the logging package out completely and "logging to
printf" while chasing particularly nasty bugs...

-t