symbol maintenance

Fabian Cenedese <Cenedeseindel.ch> writes:
>>> This is a gdb 6.5 cross-compiled with cygwin on
WinXP for ppc.
>>>
>>> (gdb) file gccext3.x
>>> Reading symbols from
/data/gdb-6.5/gdb/gccext3.x...unknown symbol type 0x1e...ex
>>> panding to full symbols...unsupported const
value attribute form: 'DW_FORM_strin
>>> g'...const value length mismatch for
'ciOffline', got 8, expected 0...done.
>>
>>Those look more serious.  GDB is trying to read
DWARF debugging
>>information, and it's malformed.  What compiler are
you using?  What
>>does 'objdump -h gccext3.x' say?
>
> The used compiler is gcc 4.0.2, cross compiled for ppc.
> The files for gccext3 were compiled with -gdwarf-2 (no
+, no O).

Oh, so you have an ELF file, not a PE file.  In that case,
could you
run 'readelf -wi gccext3.x', and post the output?  If the
DWARF in
that file is valid, and GDB's reading it properly, then the
problem is
that GCC is producing information GDB can't use.

But even in that case, GDB should be able to properly skip
data it
doesn't understand (DWARF is very good about that), so the
answer to
your original question is, no, you don't need to worry about
that
warning --- unless you're trying to use string constants and
GDB
doesn't seem to know their values.

>>>> (gdb) file gccext3.x
>>>> Reading symbols from
/data/gdb-6.5/gdb/gccext3.x...unknown symbol type 0x1e...ex
>>>> panding to full symbols...unsupported const
value attribute form: 'DW_FORM_strin
>>>> g'...const value length mismatch for
'ciOffline', got 8, expected 0...done.
>>>
>>>Those look more serious.  GDB is trying to read
DWARF debugging
>>>information, and it's malformed.  What compiler
are you using?  What
>>>does 'objdump -h gccext3.x' say?
>>
>> The used compiler is gcc 4.0.2, cross compiled for
ppc.
>> The files for gccext3 were compiled with -gdwarf-2
(no +, no O).
>
>Oh, so you have an ELF file, not a PE file.  In that
case, could you
>run 'readelf -wi gccext3.x', and post the output?  If
the DWARF in
>that file is valid, and GDB's reading it properly, then
the problem is
>that GCC is producing information GDB can't use.
>
>But even in that case, GDB should be able to properly
skip data it
>doesn't understand (DWARF is very good about that), so
the answer to
>your original question is, no, you don't need to worry
about that
>warning --- unless you're trying to use string constants
and GDB
>doesn't seem to know their values.

I uploaded the output (11MB uncompressed, 1MB zipped) here:
http://www.
speedyshare.com/512483789.html

I looked for the definition of this mismatch variable in the
source. It's:

typedef signed char int8;
const int8 ciOffline = 254;

254 for a signed char is not really nice, but that alone
shouldn't
confuse gdb or gcc, right?

 <2><6774b>: Abbrev Number: 92 (DW_TAG_variable)
     DW_AT_name        : ciOffline      
     DW_AT_decl_file   : 6      
     DW_AT_decl_line   : 142    
     DW_AT_type        : <685a0>        
     DW_AT_external    : 1      
     DW_AT_declaration : 1      
     DW_AT_const_value : 8 byte block: ff ff ff ff ff ff ff
fe  

I don't know what variable or entry would cause the
"unsupported const
value attribute form:'DW_FORM_string'", I can't find
that string.

Thanks

bye  Fabi

At 11:42 28.11.2006 -0800, Jim Blandy wrote:

>Fabian Cenedese <Cenedeseindel.ch> writes:
>>>> This is a gdb 6.5 cross-compiled with
cygwin on WinXP for ppc.
>>>>
>>>> (gdb) file gccext3.x
>>>> Reading symbols from
/data/gdb-6.5/gdb/gccext3.x...unknown symbol type 0x1e...ex
>>>> panding to full symbols...unsupported const
value attribute form: 'DW_FORM_string'...

Got it, this warning comes from a variable defined as:

const char systemTable[] = "SYSTEM";

The warning is output in dwarf2_const_value where there
really is no
case for DW_FORM_string.

bye  Fabi