[Bug bootstrap/11169] bootstrap failure after passing configure options

[rearnsha@arm.com]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11169



------- Additional Comments From rearnsha@arm.com  2003-06-12 17:25 -------
Subject: Re:  bootstrap failure after passing 
 configure options

> I'm of mixed minds about this - on the one hand, other arm-linux
> distributions use FPA; on the other, Intel defines VFP for the default
> (am I right about that?).  I suppose it wouldn't hurt to make people
> have to ask for VFP.

The linux ABI currently uses FPA: that's recognized to be a problem, given 
that FPA is not available for any "modern" processor.  ARM is currently 
working towards specifying a new EABI, and once that is complete, I'm 
hoping Linux will adopt that as its new ABI -- if they are going to 
change, better to change once than have several incompatible variants 
floating around.

If you really want to build a linux system with a different floating point 
format, then as you suggest, you can force the issue by passing -mfpu=... 
to the assembler.  I'm currently working on gcc patches to allow that (at 
present, selecting VFP word ordering on soft-float has to be done by 
hacking the config files).

R.