Re: [RFA/DWARF] constant class of DW_AT_high_pc is offset for version >=4 only.

[Mark Wielaard <mjw at redhat dot com>]

Hi Joel,

On Sat, 2014-02-15 at 19:40 +0400, Joel Brobecker wrote:
> Starting with DWARF version 4, the description of the DW_AT_high_pc
> attribute was amended to say:
> 
>    if it is of class constant, the value is an unsigned integer offset
>    which when added to the low PC gives the address of the first
>    location past the last instruction associated with the entity.
> 
> A change was made in Apr 27th, 2012 to reflect that change:
> 
>   | commit 91da14142c0171e58a91ad58a32fd010b700e761
>   | Author: Mark Wielaard <mjw@redhat.com>
>   | Date:   Fri Apr 27 18:55:19 2012 +0000
>   |
>   |     * dwarf2read.c (dwarf2_get_pc_bounds): Check DW_AT_high_pc form to
>   |     see whether it is an address or a constant offset from DW_AT_low_pc.
>   |     (dwarf2_record_block_ranges): Likewise.
>   |     (read_partial_die): Likewise.
> 
> Unfortunately, this new interpretation is now used regardless of
> the CU's DWARF version. It turns out that one of WindRiver's compilers
> (FTR: Diabdata 4.4) is generating DWARF version 2 info with
> DW_AT_high_pc attributes using the data4 form. Because of that,
> we miscompute all high PCs incorrectly.

Sorry about that. I read the spec as disallowing anything except
DW_FORM_addr to encode class address. And since before DWARF4
DW_AT_high_pc only allowed class address it hadn't occurred to me that
it could have been encoded differently. I do think this really is a bug
in the compiler that generated this.

> This patch fixes it by ammending a bit the changes from the commit
> mentioned above to use the new interpretation of a relative address
> only when the CU's DWARF version is 4 or greater.

That should work fine. GCC at least doesn't generate DW_AT_high_pc
attributes as offsets (constant class) unless it generates DWARF4+.

Cheers,

Mark