This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug bootstrap/11169] bootstrap failure after passing configure options
- From: "drow at mvista dot com" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: 12 Jun 2003 14:21:14 -0000
- Subject: [Bug bootstrap/11169] bootstrap failure after passing configure options
- References: <20030612091735.11169.jeroen.dobbelaere@acunia.com>
- Reply-to: gcc-bugzilla at gcc dot gnu dot org
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 drow@mvista.com 2003-06-12 14:21 -------
Subject: Re: bootstrap failure after passing configure options
On Thu, Jun 12, 2003 at 02:15:31PM -0000, jeroen.dobbelaere@acunia.com wrote:
> 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 jeroen.dobbelaere@acunia.com 2003-06-12 14:15 -------
> > Is this any different from building a compiler and specifying
> > -mcpu=xscale on the command line? I believe it isn't, and this sort of
> > inconsistency is _precisely_ what the patch was intended to accomplish.
> > If it is different, then I goofed.
>
> Nope, the problem is that a similar construct is not done when specifying the
> floating point unit (--with-fpu=...), although this will also need some changes
> to the 'specs' file.
Sorry, not sure what your answer to my question was...
> > Well, the Intel ABI specifies that Xscale should default to VFP,
> > doesn't it? That's the problem. If you want to build a compiler that
> > defaults to FPA instead, add support for a GCC option to specify FPA
> > (is there one already?) and add support for it to --with-float. My two
> > cents.
>
> I should check that one. If I remember correctly, on ARM it is not just a simple
> choice of -mfloat=hard/-mfloat=soft.
>
> On our platform, we try to be binary compatible with familiar (armv4), but all
> binaries are optimized for xscale. The reason I 'complain' here, is that the
> 'old behavior' suddenly changed, breaking my automated tester :(
If you want to be binary compatible to familiar, then yes, you need to use
FPA. But you should have to specify an option to get that - the Xscale
C ABI from Intel claims VFP. I'm not sure where to get a copy, sorry.