On Sun, Apr 29, 2012 at 6:37 PM, Jan Kratochvil
<jan.kratochvil@redhat.com> wrote:
> On Mon, 30 Apr 2012 02:06:13 +0200, Doug Evans wrote:
>> On Sun, Apr 29, 2012 at 2:04 AM, Jan Kratochvil <jan.kratochvil@redhat.com> wrote:
>> >  Running gdb/testsuite/gdb.ada/array_bounds.exp ...
>> >  PASS: gdb.ada/array_bounds.exp: compilation bar.adb
>> > -PASS: gdb.ada/array_bounds.exp: print itable'first
>> > +FAIL: gdb.ada/array_bounds.exp: running to bar.adb:24 in runto (GDB internal error)
>> > +ERROR: Couldn't run array_bounds/bar
> [...]
>> I can't recreate this on my fc16 box, though I see the issue.
>> What gcc are you using?
>
> Standard fc16 GCC.  But I have tested now a pre-requisite is:
> $ rpm -qf /usr/lib/debug/usr/lib64/libgnat-4.6.so.debug
> gcc-debuginfo-4.6.3-2.fc16.x86_64


I'm currently testing this patch.
Can you check that it fixes the regression for you?
TIA

[There's no longer a need to distinguish reading CUs and TUs like I
did before so I just collapsed them all into one function.]


2012-04-30  Doug Evans  <dje@google.com>

        * dwarf2read.c (init_cutu_and_read_dies): Renamed from
        init_and_read_dies_worker.  All callers updated.
        (init_cu_and_read_dies, init_tu_and_read_dies): Delete.  All calls
        replaced with init_cutu_and_read_dies.
        (load_partial_comp_unit): Pass 1 for use_existing_cu.
        (find_partial_die): Remove FIXME.  Don't free current CU.

        testsuite/
        * gdb.dwarf2/pr13961.S: New file.
        * gdb.dwarf2/pr13961.exp: New file.