Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 16 Jan 2012 13:34:27 -0500
From:      David Schultz <das@FreeBSD.ORG>
To:        Kostik Belousov <kostikbel@gmail.com>
Cc:        svn-src-head@FreeBSD.ORG, svn-src-all@FreeBSD.ORG, src-committers@FreeBSD.ORG
Subject:   Re: svn commit: r230191 - in head: lib/libc/arm/gen sys/arm/include
Message-ID:  <20120116183427.GA86151@zim.MIT.EDU>
In-Reply-To: <20120116083836.GD31224@deviant.kiev.zoral.com.ua>
References:  <201201160408.q0G48UrQ014730@svn.freebsd.org> <20120116041143.GA82129@zim.MIT.EDU> <20120116083836.GD31224@deviant.kiev.zoral.com.ua>

next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, Jan 16, 2012, Kostik Belousov wrote:
> On Sun, Jan 15, 2012 at 11:11:43PM -0500, David Schultz wrote:
> > On Mon, Jan 16, 2012, David Schultz wrote:
> > > Author: das
> > > Date: Mon Jan 16 04:08:29 2012
> > > New Revision: 230191
> > > URL: http://svn.freebsd.org/changeset/base/230191
> > > 
> > > Log:
> > >   Implement FLT_ROUNDS for arm.  Some (all?) arm FPUs lack support  for
> > >   dynamic rounding modes, but FPUless chips that use softfloat can support it
> > >   because everything is emulated anyway.  (We presently have incomplete
> > >   support for hardware FPUs.)
> > >   
> > >   Submitted by:	Ian Lepore
> > 
> > Incidentally, all of gcc's hooks into softfloat should probably be in
> > the public symbol namespace instead of FBSDprivate.  The compiler generates
> > references to them, so we cannot claim that they are internal, unsupported
> > interfaces.  I assume that moving them will not break the ABI because
> > FreeBSDprivate includes FBSD_X, but I haven't tested this.  Any objections
> > to moving them?  Affects arm and mips.
> 
> Move will break the ABI. Namespace inheritance is ignored when searching
> the symbol match.
> 
> On the other hand. FBSDprivate_1.0 is explicitely created to be changed,
> so removal of the symbols from this namespace if fine from the POV of
> the project policy.
> 
> Another argument is that both MIPS and ARM are the second-tier architectures,
> and again, project policy allows ABI breakage.

Right; it was more a question of whether it would cause anyone
undue inconvenience.

Actually, before we call them officially supported, another
question is why all of the symbols related to floating-point
emulation are coming from libc and not libgcc.



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20120116183427.GA86151>