lm32 port and gcc 4.3

[Ralf Corsepius]

Joel Sherrill wrote:
> Committed.  Thanks.
> 
> Ralf.. the patch to add lm32 support was posted to gcc-patches
> but didn't get reviewed/merged.  All the other tools have
> lm32 support in CVS/SVN now.  Can you look into how much trouble it
> would be to start building lm32 RPMs now?
I did, ... but ...

>  I don't know
> if it makes sense to use snapshot tarballs or patches.

Well, unlike to what the GCC-4.4.x release managers tell, my impression 
on GCC-trunk is
a) the code base is still very volatile (too volatile)
b) be still rather unstable (In Fedora, I am observing bugs going 
on/off/on/off nearly on a daily basis)

Also, when trying to build gcc-trunk with the lm32.patch, I noticed it 
to be incomplete (all lm32*rtems* pieces are missing, atomic ops in 
libstdc++ are missing - Is c++ supported at all?) and lm32 pieces to be 
missing and/or broken in newlib.


Additionally, I tried to back-port the lm32.patch to gcc-4.3.3.
So far, without much success.


That is, I feel we can choose between 3 evils:

a) Try to build an lm32 toolchain based on binutils-snapshots + 
gcc-snapshots.
Problems:
* Lack of a stable code basis
* significant amount of additional work on my part.
* reflect gcc-4.4.0's changes to rpm.specs (e.g. libgomp!)

b) Not to try shipping lm32 toolchains until at least gcc has stabilized
(Unlike gcc-svn, binutils-cvs is rather stable)


c) Try to back-port lm32-gcc patches to gcc-4.3.3
Problem: So far no functional patch. The gcc-4.4.0 patch is incompatble 
to gcc-4.4.0



Ralf