[fpc-devel] dwarf for "array [ 0..191] of "

[Jonas Maebe]

On 19 Jan 2013, at 15:17, Martin wrote:

> Since I don't know the fpc cdoebase, I can not check this myself.
> 
> I suspect it is a gdb issue. But if it is not to much work, maybe someone who does know fpc dwarf generation can check that it is not fpc?
> 
> I got the following source:
>    type  TFoo = array [0..191] of integer;
> 
> and in gdb
>  ptype TFoo
> 
> Using dwarf, up to gdb 7.2 (inclusive) the result is ok.
> 
> but 7.3 up reports
>    array [0..-65] of LONGINT
> 
> SO definitely something changed in GDB.
> But I do not know if the old or new gdb is right.

It looks like a gdb issue (from dwarfdump; readelf -g can show similar information):

0x00000076:     TAG_array_type [4] *
                 AT_name( "TFOO" )
                 AT_byte_size( 0x00000180 )
                 AT_type( {0x00000091} ( SMALLINT ) )

0x00000082:         TAG_subrange_type [5]  
                     AT_lower_bound( 0x0000000000000000 )
                     AT_upper_bound( 0x00000000000000bf )
                     AT_byte_stride( 0x00000002 )
                     AT_type( {0x000000b0} ( BYTE ) )

0x0000008b:         NULL

...
                 AT_name( "BYTE" )
                 AT_type( {0x000000ba} ( BYTE ) )

0x000000ba:     TAG_base_type [7]  
                 AT_name( "BYTE" )
                 AT_encoding( DW_ATE_unsigned )
                 AT_byte_size( 0x01 )

I.e., FPC declares the range of the array as BYTE, and BYTE is declared as an unsigned base type of 1 byte large.


Jonas
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freepascal.org/pipermail/fpc-devel/attachments/20130120/d33718f8/attachment.html>