using dwarf2
var s: string[5]
will generate info as shown at the end of this mail
The array "st" is at
DW_AT_data_member_location: 2 byte block: 23 1 (DW_OP_plus_uconst: 1)
That means, the array does not contain the length.
Yet the bounds of the array are 0 to 5.
That is one to much. It also leads to gdb showing the wrong char for
"s[1]". Should show first, but does show 2nd
While on the subject. Is there any better way to encode ansistrings?
(currently simply as pointer to char). I hoped it was only gdb
shortcoming that they are indistinguishable from pchar (0 based idx vs
1 based). But actually fpc, encodes them as the same thing. (even dwarf2
has DW_TAG_string_type / I dont know about gdb...) DW_TAG_string_type
allows even for a length. It does not have a start index...
<2><1b5>: Abbrev Number: 6 (DW_TAG_structure_type)
<1b6> DW_AT_name : ShortString
<1c2> DW_AT_byte_size : 6
<3><1c3>: Abbrev Number: 7 (DW_TAG_member)
<1c4> DW_AT_name : length
<1cb> DW_AT_data_member_location: 2 byte block: 23 0
(DW_OP_plus_uconst: 0)
<1ce> DW_AT_type : <0x230>
<3><1d2>: Abbrev Number: 7 (DW_TAG_member)
<1d3> DW_AT_name : st
<1d6> DW_AT_data_member_location: 2 byte block: 23 1
(DW_OP_plus_uconst: 1)
<1d9> DW_AT_type : <0x1de>
<2><1de>: Abbrev Number: 8 (DW_TAG_array_type)
<1df> DW_AT_byte_size : 6
<1e0> DW_AT_stride : 1
<1e1> DW_AT_type : <0x2a0>
<3><1e5>: Abbrev Number: 9 (DW_TAG_subrange_type)
<1e6> DW_AT_lower_bound : 0
<1e7> DW_AT_upper_bound : 5
<1e8> DW_AT_type : <0x230>
_______________________________________________
fpc-devel maillist - [email protected]
http://lists.freepascal.org/mailman/listinfo/fpc-devel
