Re: Inconsistent DWARF dump behavior between readelf and objdump

[Joerg Sonnenberger]

On Mon, Jul 17, 2017 at 10:14:01AM +0930, Alan Modra wrote:
> On Sun, Jul 16, 2017 at 06:37:09PM +0200, Joerg Sonnenberger wrote:
> > Hello,
> > When assembling the following input with the binutils 2.27:
> > 
> >     .text
> >     .cfi_startproc
> >     .cfi_adjust_cfa_offset 64
> >     .skip 75040, 0
> >     .cfi_adjust_cfa_offset 128
> >     .cfi_endproc
> > 
> > I get the attached object file. readelf --debug-dump=frames gives:
> > 
> >     00000018 00000018 0000001c FDE cie=00000000 pc=00000000..00012520
> >       DW_CFA_advance_loc4: 75040 to 00012520
> >       DW_CFA_def_cfa_offset: 192
> >     ...
> > 
> > and objdump -Wf gives:
> > 
> >     00000018 0000000000000018 0000001c FDE cie=00000000 
> > pc=0001252000000000..c00f252125200400
> >       Augmentation data:     00
> > 
> >       DW_CFA_nop
> >       DW_CFA_nop
> 
> I could not reproduce anything like this using binutils-2.27 or 2.26
> on x86_64-linux.

Certainly reproducable with 2.27 (original try) and 2.26 on Linux. But
it seems like a freshly build 2.28 does not have this behavior.

Joerg