Re: [RFA] Repost ARM frame patches

[Richard Earnshaw <rearnsha at arm dot com>]

> On Tue, Sep 02, 2003 at 11:52:26PM +0100, Richard Earnshaw wrote:
> > Daniel,
> > 
> > My appologies for not reviewing this before.  Just way too much day job...
> > 
> > It looks fine to me (at least, it seems better than what's there at 
> > present).
> > 
> > My only question is, once we start using the dwarf2 unwinder, can it cope 
> > with the fact that gcc currently does not emit frame unwind information 
> > for Thumb code? (ie can it handle a mix of code that uses dwarf2 and 
> > traditional unwinding?)
> 
> I was going to say yes, but...
> 
> drow@nevyn:/big/fsf/projects/arm/obj/gdb/testsuite/interwork% readelf -wf test1.o
> The section .debug_frame contains:
> 
> 00000000 0000000c ffffffff CIE
>   Version:               1
>   Augmentation:          ""
>   Code alignment factor: 1
>   Data alignment factor: -4
>   Return address column: 14
> 
>   DW_CFA_def_cfa: r13 ofs 0
> 
> 00000010 0000000c 00000000 FDE cie=00000000 pc=00000000..0000001c
> 
> 00000020 0000000c 00000000 FDE cie=00000000 pc=0000001c..00000046
> 
> 
> i.e. GCC emits _empty_ dwarf unwind information for thumb functions,
> rather than none at all.  That's unlikely to work.  We'd need to modify
> the dwarf2 unwinder to ignore empty FDEs.
> 
> I'll check in the non-dwarf parts now, and then we can figure out what
> to do about that.

Though of course a trivial leaf function *will* have an empty FDE.  
Consider

int func(void) { return 0;}

Which compiles to 

	mov	r0, #0
	bx	lr

I would have thought that wouldn't need any frame unwind information.  So 
we would have a problem distinguishing trivial cases from "not generated" 
cases.

R.