This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: .cfi_sections


On Tue, Dec 08, 2009 at 12:30:15PM -0800, Ulrich Drepper wrote:
> On 12/08/2009 10:50 AM, Thomas Schwinge wrote:
> >I have a need for using .cfi_sections in glibc code.  .cfi_sections is a
> >recent addition to binutils' as, done by Jakub on 2009-10-02.
> 
> If you have a need, then this is absolute.  cfi directs codeflow and
> if you don't have the feature the results are not the same.  If this
> is the case you simply have to demand for that platform binutils
> which are recent enough.

I think the fallback is OK in this case:

* Either way, generate ARM (non-DWARF) unwind tables.  We already do
this.  Runtime always uses these.

* If .cfi_sections is available, use it to generate only .debug_frame.
Include CFI for debugger use.  GDB can't use the ARM tables.

* If .cfi_sections is not available, use CFI anyway.  It will generate
.eh_frame by default.  Nothing at runtime will read .eh_frame so that
wastes some read-only text space.  GDB still works.

Today, GDB doesn't work.

-- 
Daniel Jacobowitz
CodeSourcery


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]