Re: The root cause for SEGV in evaluating fortran function call, any solution or suggestion?

[Wu Zhou <woodzltc@cn.ibm.com>]

Hi Mark,

On Wed, 2 Nov 2005, Mark Kettenis wrote:

> What you describe sounds very similar to what
> gdbarch_stabs_argument_has_addr is all about.  Currently these are
> only used on SPARC, and only for stabs.

Thanks for your pointer.  It looks very much like the same (If I understand
correctly :-).  Look at the following testcase please:

        PROGRAM funcall
          integer a, b
          b = 2
          a = res(b)
        END PROGRAM

        function res(m)
          integer m
          integer res
          res = m * m
        end function

The debuginfo for function res is like this:

 <1><da>: Abbrev Number: 7 (DW_TAG_subprogram)
     DW_AT_sibling     : <10f>
     DW_AT_external    : 1
     DW_AT_name        : res_
     DW_AT_decl_file   : 1
     DW_AT_decl_line   : 23
     DW_AT_type        : <bb>
     DW_AT_low_pc      : 0x804863f
     DW_AT_high_pc     : 0x8048655
     DW_AT_frame_base  : 1 byte block: 55       (DW_OP_reg5)
 <2><f5>: Abbrev Number: 8 (DW_TAG_formal_parameter)
     DW_AT_name        : m
     DW_AT_type        : <10f>   =====> This is a const pointer to integer
     DW_AT_artificial  : 1
     DW_AT_location    : 2 byte block: 91 8     (DW_OP_fbreg: 8)

 <1><10f>: Abbrev Number: 9 (DW_TAG_const_type)
     DW_AT_type        : <114>
 <1><114>: Abbrev Number: 10 (DW_TAG_pointer_type)
     DW_AT_byte_size   : 4
     DW_AT_type        : <bb>
 <1><bb>: Abbrev Number: 6 (DW_TAG_base_type)
     DW_AT_name        : integer
     DW_AT_byte_size   : 4
     DW_AT_encoding    : 5      (signed)

So my question is: how will gdb to handle this situation if the gdbarch 
shows that stabs argument has addr, especially when the command line pass 
an integer type, but the debuginfo said that it need a pointer?

> 
> One could argue that the debug information generated by g77 is wrong,
> because it doesn't reflect the actual implementation of FUNC_NAME.  Or
> perhaps GDB symbol reading code causes problems.  Can you post a
> concrete example of a function call that goes wrong, and add a bit of
> explanation about the types involved for those of us who are not very
> familiar with Fortran?

Both of your arguments seem reasonable.  But even in the first argument it 
might also be desirable that we could handle it gracefully, right? 

Here is the debugging session for the above testcase (maybe I could 
convert it into a real testcase for gdb testsuite?), hoping that 
it could clarify the things a little.

(gdb) r
Starting program: /root/DE/gdb-6.3/gdb/testsuite/gdb.fortran/func-call

Breakpoint 1, MAIN__ () at func-call.f:19
19                b = 2
Current language:  auto; currently fortran
(gdb) n
20                a = res(b)
(gdb) p res (2)
No symbol "res" in current context.
(gdb) p res_ (2)     =========> Use constand '2' as argument, get SEGV

Program received signal SIGSEGV, Segmentation fault.
0x08048648 in res_ (m=0x2) at func-call.f:26
26                res = m * m
The program being debugged was signaled while in a function called from 
GDB.
GDB remains in the frame where the signal was received.
To change this behavior use "set unwindonsignal on"
Evaluation of the expression containing the function (res_) will be 
abandoned.
(gdb) r
The program being debugged has been started already.
Start it from the beginning? (y or n) y

Starting program: /root/DE/gdb-6.3/gdb/testsuite/gdb.fortran/func-call

Breakpoint 1, MAIN__ () at func-call.f:19
19                b = 2
(gdb) n
20                a = res(b)
(gdb) p res_ (b)    =========> Use variable 'b' as argument, get SEGV

Program received signal SIGSEGV, Segmentation fault.
0x08048648 in res_ (m=0x2) at func-call.f:26
26                res = m * m
The program being debugged was signaled while in a function called from 
GDB.
GDB remains in the frame where the signal was received.
To change this behavior use "set unwindonsignal on"
Evaluation of the expression containing the function (res_) will be 
abandoned.
(gdb) r
The program being debugged has been started already.
Start it from the beginning? (y or n) y

Starting program: /root/DE/gdb-6.3/gdb/testsuite/gdb.fortran/func-call

Breakpoint 1, MAIN__ () at func-call.f:19
19                b = 2
(gdb) n
20                a = res(b)
(gdb) p res_ (&b)    ========> Use '&b' as argument, get correct result
$1 = 4

Regards
- Wu Zhou