Re: RFA: Support for Thumb in dynamic objects

[Richard Earnshaw <rearnsha@gcc.gnu.org>]

On Wed, 2004-11-17 at 17:05, Daniel Jacobowitz wrote:
> On Wed, Nov 17, 2004 at 04:59:10PM +0000, Richard Earnshaw wrote:
> > On Wed, 2004-11-17 at 16:48, Daniel Jacobowitz wrote:
> > > On Wed, Nov 17, 2004 at 01:48:06PM +0000, Richard Earnshaw wrote:
> > > > > OK?  Comments?
> > > > 
> > > > This is OK once you've addressed the point Paul raised.  You might also
> > > > have to look at the tests when run in big-endian mode too.
> > > 
> > > I missed some arm-elf vs arm-linux issues, and some big vs little
> > > endian issues; no one's run the testsuite in big-endian in a while.
> > > I'm reposting for review.  There are two changes in the patch that I
> > > would like someone else to look at:
> > > 
> > >   - I fixed a big-endian Thumb disassembly bug.  It would read past
> > >     the end of the section.
> > 
> > Can you use a macro rather than ~0x3?
> 
> I just copied it from a couple of lines up.  How about:
>   ROUND_DOWN (pc + 4, 4)
> instead?

It would certainly be better than the direct manipulation, but I was
really thinking 'why are we doing the masking at this point?'  Could it
be because we want an ARM instruction address?  If it's a Thumb insn
address why isn't it ~1?  A suitably named macro would convey that
information directly...

R.