Re: [tree-ssa] Mainline merge plan

[Richard Earnshaw <rearnsha at arm dot com>]

> On Tue, 2004-02-24 at 13:10, Richard Earnshaw wrote:
> 
> > I hope the GMP dependencies I reported in an earlier message are also 
> > going to be sorted out before the merge.
> > 
> Point me to a URL?  I vaguely recall a GMP discussion, but not what the
> consensus was.
> 

http://gcc.gnu.org/ml/gcc/2004-02/msg00742.html

lists a number of problems building on targets that don't have gmp 
installed in as a system library.

> > I'd also like to see arm-elf building as well (you have no ARM target in 
> > your list).
> > 
> Would you be willing to test arm-elf?
> 

I'm happy to help out with testing, but I don't have enough bandwidth to 
do much more than fix trivial issues that are ARM-specific.

> Also, remember that at the time of the merge, mainline will still be in
> stage 1.  We do not expect a release ready compiler by the time we do
> the merge.  Those problems will be addressed post-merge.
> 

And?  That doesn't give you carte-blanche to cause wholesale breakage.  
Please remember that other people are trying to develop GCC, and to have 
the whole world pulled out from under their feet is very disruptive.

Anyway, here's some items I found last night:

The ARM compiler doesn't build at all -- RTH's change to nested function 
support has left arm.c uncompilable -- the changes are (I think) pretty 
trivial, but there are so many other problems at this time that I can't 
even test what I think is the right change.  That patch broke *at least* 
two back ends.

The fortran compiler doesn't run on the systems I have -- see the GMP 
thread above.

The build system tries to build libmudflap with pthread calls.  An arm-elf 
target is never going to support pthreads (it's for bare-metal systems).  
Consequently it's impossible to build the libjava or objc libraries.

R.