Return from manually called function in gdb

Orjan Friberg orjan.friberg@axis.com
Mon Sep 11 07:17:00 GMT 2000


"Peter.Schauer" wrote:
> 
> It is amazing that no one noticed this before.
> Please try the following patch and let me know if it works for you:

For native (i686-pc-linux-gnu) gdb it seems to work fine. I put a
breakpoint at various functions (taking primitive types as well as
strings and structs as parameters), called them, and forced a return.
Diff'ing the register contents before and after the function call told
me the registers had not changed. (A DejaGnu-related note: there is only
one return command issued in the gdb.base testsuite, and the function
returned from is not called by the test case, but by the code.)

The problem still persists with my gdb-port, but that's another story :)

Thanks,
Orjan

> 
> *** gdb/stack.c.orig    Sat Aug  5 11:37:49 2000
> --- gdb/stack.c Mon Sep 11 11:36:11 2000
> ***************
> *** 1886,1891 ****
> --- 1886,1897 ----
> 
>     POP_FRAME;
> 
> +   /* If we are at the end of a call dummy now, pop the dummy frame too.  */
> +
> +   if (CALL_DUMMY_HAS_COMPLETED (read_pc(), read_sp (),
> +                               FRAME_FP (get_current_frame ())))
> +     POP_FRAME;
> +
>     /* Compute the return value (if any) and store in the place
>        for return values.  */
> 
> 
> > Hi,
> >
> > This problem occurs on native (in my case i686-pc-linux(-gnu)) on both
> > the 2000-05-19 release, as well as the latest CVS sources. What happens
> > is that when I set a breakpoint at a function, call it from gdb, and
> > force a return, gdb is thoroughly confused upon return. Allow me to
> > illustrate what happens (using the funcargs test case from gdb.base):
> >
> >
> > (gdb) b add
> > Breakpoint 2 at 0x8048443: file
> > ../../../src/gdb/testsuite/gdb.base/callfuncs.c, line 58.
> > (gdb) call add(4,5)
> >
> > Breakpoint 2, add (a=4, b=5)
> >     at ../../../src/gdb/testsuite/gdb.base/callfuncs.c:58
> > 58        return (a + b);
> > The program being debugged stopped while in a function called from GDB.
> > When the function (add) is done executing, GDB will silently
> > stop (instead of continuing to evaluate the expression containing
> > the function call).
> > (gdb) return
> > Make add return now? (y or n) y
> > #0  <function called from gdb>
> > (gdb) bt
> > #0  <function called from gdb>
> > #1  main () at ../../../src/gdb/testsuite/gdb.base/callfuncs.c:185
> > #2  0x4004ca42 in __libc_start_main () from /lib/libc.so.6
> > (gdb) n
> > warning: Cannot insert breakpoint 0:
> > Cannot access memory at address 0x4
> >
> >
> > I'm merely using native gdb as a reference for my gdb port (which is for
> > cross-debugging), and I'm not very familiar with the i386 registers.
> > What happens with my gdb port is that the dummy pc used to recognize
> > that we are in a call dummy is written to the pc when continuing from
> > the breakpoint. Regular function calls from within gdb, when there is no
> > breakpoint in the called function, works fine.
> >
> > I'm probably asking for trouble by forcing a return in that situation,
> > but I don't think the situation at hand is unimaginable.
> >
> >
> > Orjan Friberg
> > Axis Communications AB
> > E-mail: orjan.friberg@axis.com
> > Phone:  +46 46 272 17 68
> 
> --
> Peter Schauer                   pes@regent.e-technik.tu-muenchen.de


More information about the Gdb mailing list