[uClibc]Re: COFF vs. ELF on m68k

James Graves jgraves at deltamobile.com
Sun Mar 4 07:27:28 UTC 2001


On Wed, 28 Feb 2001, David McCullough wrote:
> Jivin Erik Andersen lays it down ...
> > Well, I'm not much of an m68k person myself, but that patch was submitted
> > by a person getting internal compiler errors on const stuff...  Granted
> > this is a lame way to solve it (and I said as much at the time) but since
> > it hasn't bothered me, I've left it alone.

That would be me.  :-)

> None of the non-mmu CPU's can do fork,  but I think you can leave the
> VFORK_MACRO stuff until someone (probably me :-) gets around to adding the
> vfork system call as a function and not a macro.
> 
> I think fork will still work even with VFORK_MACRO define.  I may be wrong.
> 
> I actually just removed the const things,  don't think I was checking it in
> though.  This stuff should be in Config and not in Rules.mak IMHO.  That
> way you can override it or do whatever you want.

Yes, I guess this should definitely be in the Config file instead.
Currently, the Config file doesn't have a variable to specify
additional stuff for CFLAGS, should we add one?



BTW, to get the current CVS version of uC-libc to compile for
m68k-pic-coff, I had to do:

1) Edit the Config.m68k file (obviously).

2) Comment out all the .weak references because the 2.7.2 compiler
doesn't like them.  (I'm not enough of a compiler guru to fix this.)

3) Comment out the C++ initialization stuff in crt0.S.  Apparently,
the compiler saw this and tried add in some routines to setup C++
support (I guess), but this broke badly.

I can send someone diffs if they're still running the old toolchain.

Thanks for all your work everyone,

James Graves
Delta Mobile Software






More information about the uClibc mailing list