Re: [PATCH] Fix debug info after the fortran descriptor changes (PR debug/84131)

[Steve Kargl]

On Tue, Jan 30, 2018 at 11:53:40PM +0100, Jakub Jelinek wrote:
> 
> DW_AT_data_location needs to be the data pointer, which is at offset 0,
> but we were emitting a bogus +8 in that case (+4 for 32-bit targets).
> 
> With this patch, the change on -g -dA is:
> -   .uleb128 0x4# DW_AT_data_location
> +   .uleb128 0x2# DW_AT_data_location
> .byte   0x97# DW_OP_push_object_address
> -   .byte   0x23# DW_OP_plus_uconst
> -   .uleb128 0x8
> .byte   0x6 # DW_OP_deref
> and gdb can debug the array in there again.
> 
> Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
> 

Yes.  Thanks for looking at this!

-- 
Steve

[PATCH] Fix debug info after the fortran descriptor changes (PR debug/84131)

[Jakub Jelinek]

Hi!

DW_AT_data_location needs to be the data pointer, which is at offset 0,
but we were emitting a bogus +8 in that case (+4 for 32-bit targets).

With this patch, the change on -g -dA is:
-   .uleb128 0x4# DW_AT_data_location
+   .uleb128 0x2# DW_AT_data_location
.byte   0x97# DW_OP_push_object_address
-   .byte   0x23# DW_OP_plus_uconst
-   .uleb128 0x8
.byte   0x6 # DW_OP_deref
and gdb can debug the array in there again.

Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?

2018-01-30  Jakub Jelinek  

PR debug/84131
* trans-array.c (gfc_get_descriptor_offsets_for_info): Set *data_off
to DATA_FIELD's offset rather than OFFSET_FIELD's offset.

--- gcc/fortran/trans-array.c.jj2018-01-26 12:43:25.164922494 +0100
+++ gcc/fortran/trans-array.c   2018-01-30 19:34:01.844232363 +0100
@@ -511,7 +511,7 @@ gfc_get_descriptor_offsets_for_info (con
   tree type;
 
   type = TYPE_MAIN_VARIANT (desc_type);
-  field = gfc_advance_chain (TYPE_FIELDS (type), OFFSET_FIELD);
+  field = gfc_advance_chain (TYPE_FIELDS (type), DATA_FIELD);
   *data_off = byte_position (field);
   field = gfc_advance_chain (TYPE_FIELDS (type), DTYPE_FIELD);
   *dtype_off = byte_position (field);

Jakub