[RFA/alpha] Fix problem loading core files

Joel Brobecker brobecker@adacore.com
Wed Dec 1 04:31:00 GMT 2004

I noticed that GDB is no longer able to load the registers from a core

        (gdb) core core
 OK  -> Core was generated by `crash'.
 OK  -> Program terminated with signal 6, Aborted.
 !?! -> Register  not found in core file.

That lead me to a couple of issues. First, in fetch_osf_core_registers()
in alpha-nat.c; we have an array named core_reg_mapping[ALPHA_NUM_REGS]
which value is statically defined (minus a #if ... #else ... #endif

  static int const core_reg_mapping[ALPHA_NUM_REGS] =
#ifdef NCF_REGS
#define EFL NCF_REGS
    CF_V0, CF_T0, CF_T1, CF_T2, CF_T3, CF_T4, CF_T5, CF_T6,

The problem is that ALPHA_NUM_REGS has been incremented on 2003-01-31:

        * alpha-nat.c (REGISTER_PTRACE_ADDR): Merge into ...
        (register_addr): ... here.  Support ALPHA_UNIQUE_REGNUM.
        (fetch_elf_core_registers): Support ALPHA_UNIQUE_REGNUM.
        * alpha-tdep.c (alpha_register_name): Add "unique".
        * alpha-tdep.h (ALPHA_NUM_REGS): Increment.
        * config/alpha/nm-linux.h (ALPHA_UNIQUE_PTRACE_ADDR): New.

But the table was not updated in consequence. So I added an index
of -1 for this extra register in the table. So the number of elements
in the value matches the number of elements declared.

But that was not the problem that caused the error above. What happens
is that we removed a vitual register on 2003-06-01:

        * alpha-tdep.h (ALPHA_FP_REGNUM): Remove.
        * alpha-tdep.c (alpha_register_name): Remove vfp entry.
        (alpha_cannot_fetch_register): Remove ALPHA_FP_REGNUM.
        (alpha_cannot_store_register): Likewise.
        * alphabsd-nat.c (fetch_inferior_registers): Don't set FP_REGNUM.
        * alpha-nat.c (supply_gregset): Likewise.
        * alphanbsd-tdep.c (fetch_core_registers): Likewise.

The register index was kept reserved to maintain compatibility with
alpha remote debug agents. But the problem is that now
alpha_cannot_fetch_register() considers this old register as
fetchable, which is not true. As a consequence,
fetch_osf_core_registers() finds an inconsistency because on
the one hand it has a negative offset (-1), and on the other
hand it has a register that presumably is fetchable.

And hence we get the error.

This patch modified alpha_cannot_fetch_register() and
alpha_cannot_store_register() to exclude the register which
have an empty name. These empty names are the sign of a register
number that is no longer used, but has been reserved to maintain
compatibility with remote debug agents.

2004-11-30  Joel Brobecker  <brobecker@gnat.com>

        * alpha-tdep.c (alpha_register_name): Add comment.
        (alpha_cannot_fetch_register): Exclude register number which
        are no longer used from the list of registers that can be fetched.
        (alpha_cannot_store_register): Exclude register number which
        are no longer used from the list of registers that can be stored.

Tested on alpha-tru64. Makes the following changes to the testsuite

* gdb.base:
|       FAIL | PASS       | corefile.exp: args: -core=corefile                 |
|       FAIL | PASS       | corefile.exp: args: execfile -core=corefile        |
|       FAIL | PASS       | corefile.exp: core-file command                    |
|       FAIL | PASS       | corefile.exp: print func2::coremaker_local         |
|       FAIL | PASS       | corefile.exp: backtrace in corefile.exp            |

OK to apply?


More information about the Gdb-patches mailing list