[Bug gdb/12130] gdb: backtrace on stripped binary much slower thanon binary with debug info

["edwin+sourceware at etorok dot net"<sourceware-bugzilla at sourceware dot org>]

http://sourceware.org/bugzilla/show_bug.cgi?id=12130

--- Comment #5 from TÃrÃk Edwin <edwin+sourceware at etorok dot net> 2012-07-13 19:55:58 UTC ---
(In reply to comment #4)
> (In reply to comment #3)
> > With current CVS trunk gdb I get pretty similar results for both:
> > 
> > barimba. sh -x RunIt.sh
> > + gcc -g -o pr pr.c
> > + echo 'With debuginfo:'
> > With debuginfo:
> > + gdb --batch-silent -ex run -ex 'bt -16' ./pr
> > 
> > real    0m21.787s
> > user    0m18.759s
> > sys    0m2.440s
> > + echo 'Without debug info (stripped):'
> > Without debug info (stripped):
> > + strip pr
> > + gdb --batch-silent -ex run -ex 'bt -16' ./pr
> > 
> > real    0m19.851s
> > user    0m17.299s
> > sys    0m2.340s
> 
> Sounds promising.
> 
> > Could you possibly try a newer gdb and see if the problem has been fixed
> > for you as well?
> 
> I get 14s vs 49s using GDB 7.4.1, will try CVS version a bit later.

With 7.4.50.20120713-cvs I get ~18s for both, so this bug is fixed, thanks!

The speed for the with-debuginfo case has regressed though:
I get ~11s with GDB 7.2, ~14s with 7.4.1 and ~18s with CVS trunk.
Feel free to close this bug though if this is to be expected.

Did some bisecting, looking for the jump from ~13s to ~16s:

This is a good commit (time ~13s for both with&without debuginfo):
Author: Jan Kratochvil <jan.kratochvil@redhat.com>  2011-10-09 22:23:41
 gdb/
        Tail call sites reader implementation.
        * dwarf2read.c (read_call_site_scope): Recognize DW_AT_GNU_tail_call,
        fill in TYPE_TAIL_CALL_LIST.
        * gdbtypes.h (struct func_type): New field tail_call_list.
        (struct call_site): New field tail_call_next.

Possible bad-commit range up to:
Author: Jan Kratochvil <jan.kratochvil@redhat.com>  2011-10-09 22:46:04
    gdb/
        Entry values NEWS entries, DWARF disassembly support.
        * NEWS: New entry values entry.
        (set print entry-values, show print entry-values)
        (set debug entry-values, show debug entry-values): New entries.
        * dwarf2loc.c (disassemble_dwarf_expression): New parameters start and
        indent.  Remove variable start.  Move header printing out.  Respect
        INDENT.  Support DW_OP_GNU_entry_value.
        (locexpr_describe_location_1): Move the header printing here, extend
        the disassemble_dwarf_expression passed parameters.

    gdb/testsuite/
        * gdb.arch/amd64-entry-value.s: New file.
        * gdb.mi/mi2-amd64-entry-value.s: New file.

        (TYPE_TAIL_CALL_LIST): New definition.

-- 
Configure bugmail: http://sourceware.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.