Re: [avr-chat] Can't Install avr-gcc in FreeBSD

[David Brown]

Joerg Wunsch wrote:
> David Brown <address@hidden> wrote:
>
> > For some port developers, the requirement of assigning copyright to
> > the FSF is too much (this could be one of the reasons why Altera,
> > Xilinx, and Microchip keep their ports outside the tree, but it's
> > unlikely to be a reason for the msp430 people since the msp430
> > binutils port is inside the FSF tree).
>
> I agree about commercial entities here.  Personally, I think it's one
> of the most arguable policies of the FSF, and it even runs into
> curious situations.  For example, I signed the copyright assignment
> for GCC, binutils, and GDB since the FSF insists on it, yet this
> assignment is completely void.  According to our German rules, the
> copyright is inherently bound to the person creating something, and
> cannot be transferred at all.  (It can be inherited only.)

I wonder if the same applies in Norwegian law?

I can understand the FSF position (the theory is that it makes it easier 
to handle abuses if the code is all owned by one entity), but I'm sure 
it is a hinder to contribution for many people.  It doesn't make much 
difference to collaboration projects like gcc, where individual's parts 
can't stand on their own.  But for many other kinds of projects, owning 
the copyrights means you can re-use the code in other projects, or 
dual-license the code (such as is done with FreeRTOS or MySQL, for example).

> OTOH, even a commercial entity developing parts of GCC is bound to the
> rules of the GPL anyway, so it's kinda questionable as well to not go
> the final step, and hand everything over to the tree.

You lose a fair amount of control when you give away your copyrights, 
and I can understand why companies would be sceptical to that.  For gcc, 
it may have little relevance since the code for the port is of little 
use outside of gcc, but corporate lawyers tend to be a sceptical bunch.

The other reason I'd guess for commercial companies keeping their ports 
outside the tree is that it leaves them free to control their own 
schedules concerning releases - Altera, for example, likes to combine 
new releases of gcc with new releases of the Nios II soft processor, and 
new releases of their FPGA toolset.  This way, they can make the changes 
when they want, and don't have to follow the FSF gcc releases.

> > I see many open source projects moving from cygwin to mingw for
> > precisely this reason (mingw has less run-time overheads, which is a
> > bonus), and this progress has made my life much easier.
>
> Well, for sure, everything that can run under MinGW should be used
> that way.  After all, it's then using the native Win32 API, rather
> than an emulated Posix API.
>
> Unfortunately, for some programs (like AVaRICE), this would
> essentially affect their entire operation, and thus require a rewrite.
> Other programs (like AVRDUDE) share a good common part of
> API-independent code, so they have been ported to the Win32 API years
> ago.

It's not that long ago since it was a hard job compiling gcc under mingw 
(last time I did it for the ColdFire was using patches from avr-gcc - 
the beauty of open source in action!).  ./configure scripts are 
sometimes a bit difficult in msys, although that too has got better in 
recent times.  But there are plenty of programs that use "fork", and 
that means compilation under cygwin if you don't want to do a lot of 
patching.

mvh.,

David