DWARF Standard

Section 3.1.1, pg 66

Hi all,

here is my proposal to extend DW_TAG_compile_unit entry with a set of attributes.
 Actually one attribute is mostly needed, namely DW_TAG_hash_value. This attribute
 is presumed to be optional. If a compile unit (DW_TAG_compile_unit) has got 
this attribute (DW_TAG_hash_value), the attibute value is generated as a hash 
function taken over all the compile unit DWARF debug information. Hash 
function might be selected as SHA1.

The following changes to DWARF5 standard document are proposed in details:

1) Add to "3.1.1 Full and Partial Compilation Unit Entries" section, the 
description of new "DW_TAG_hash_value" attribute. It is going to be 
something like (on page 66): "17. A DW_TAG_hash_value attribute whose 
value is 20 bytes SHA1 hash function (see, RFC 3174) taken over all 
the current compile unit DWARF debug information items. It means that 
if any of the current compile unit item is changes (either its attributes 
or any attribute value) the value of DW_TAG_hash_value attribute will be 
changed respectfully."

2) Add to "2.2 Attribute Types" section the description of 
"DW_TAG_hash_value" attribute: "Hash value of CU debug data". 

3) Add to "7.5.4 Attribute Encodings" section, 

   Attribute name                        |   Value   |  Classes
   -------------------------------------------------------------
   DW_TAG_hash_value                     |   0x8d    |  block

4) Add to "Appendix F. Split DWARF Object Files (Informative)" section 
the appropriate information:
                                    |             Unit Kind            
   Attribute                        | Conventional        |       Skeleton and Split
                                    | Full &       Type   | Skeleton     Split Fill    Split Type
                                    | Partial             | 
   ------------------------------------------------------------------------------------------------
   DW_TAG_hash_value                |    X                |                   X


Rationale:
The main idea behind introducing "DW_TAG_hash_value" attribute for a 
compilation unit (CU) is to reduce the time which is required by DWARF 
information parsers. And if a particular CU has been previously parsed by 
debugger (or it might be some other application), every next run the debugger 
might just check either "DW_TAG_hash_value" attribute value is changed. 
Whenever the attribute value is not changed (i.e. SHA1 hash is the same), 
it means that all the previous results from parsing CU DWARF data might 
be used again and re-parsing is not necessary. Here, different application 
might use caching techniques to make work with large DWARF debug data more 
efficient.

--

2021-10-04:  Rejected.  Additional implementation experience or proof-of
concept is needed.