Re: arm_addr_bits_remove

[Pedro Alves <pedro@codesourcery.com>]

Daniel Jacobowitz wrote:
> 
> If you're daring enough, this is OK - this could affect any target
> defining gdbarch_addr_bits_remove so keep an eye out in case hppa,
> m88k, mips, or s390 break.  Please wait another day before checking it
> in, in case someone else knows more about it.
> 

It seems possible that the hppa port would be affected.  It has has a
comment mentioning symbol tables having special bits set and
other compilers other than gcc.  From the comments, I believe
the other ports would be safe -- the extra bits should be set
at runtime only.  Other than arm, only mips choses what to strip
based on some expression, but in that case, the I believe the
extra bits won't be set on the symbol tables/line info.

> I think the original patch should be committed too.  Jim objected in
> terms of "returning the wrong answer", but that's not really the case.
> The 0x2 bit is never an extra piece of information about an address,
> like the 0x1 bit is.  It's part of the address; just if it happens
> to be part of an ARM address, executing code there is unpredictable.

Since Jim didn't really object, I'll go with the original
arm-tdep.c patch.

-- 
Pedro Alves