> On Tue, 18 Aug 1998, Richard Earnshaw wrote:
> 
> : The -msoft-float default should now be fixed in the Cygnus egcs-1.1 branch.
> 
> It's not. I imported the supposedly fixed sources, and I've been boggling
> over them; it "looks" fixed but it's still physically functioning with
> -mhard-float as default.

I'll try to look into this, but I'm running out of time before I go on 
leave.

> 
> : Which incorrectly calculates the address of foo.  I've established that 
> : the code generated by the compiler for the above is semantically the same 
> : as that generated by the old compiler, but I haven't got further than that 
> : yet.
> 
> Hm.  I'll look into it.  Of course, gcc doesn't "really" support PIC on
> ARM; we've just commented out the "PIC code not supported; ignored" warning.

1) It is a lie that the PIC flag is ignored.  It does cause PIC code to be 
generated, it just may not work.  Having said that, I'm not aware of any 
problems with the compiler part of the process.

2) The message really appears because the only version of gas that can 
currently handle PIC is the one in the netbsd source tree.  It is really 
meant as a warning that if you use this option and run into problems then 
I wouldn't be at all surprised, since the code is still fairly new and may 
have bugs in it.   A similar warning appeared on the HP port for years.

3) I put the warning in when I developed the original PIC code generator 
for the arm.  At the time I didn't have a machine I could test it on (ie 
one running netbsd), and some fixes went into the netbsd port of gcc 2.7 
that took a long time to make it back into the mainline sources.


Richard.