[Dwarf-discuss] What values to use for our own attributes

Michael Eager eager
Fri Mar 10 08:30:54 PST 2006

I talked with Ian Murdock, the new CTO for Free Standards Group, just
yesterday.  He showed me a website that they have in development for
the LSB which includes a Wiki and some reasonable content management
support.  We talked about converting the Dwarf website over to this
new style.  It would seem that a Wiki for registering vendor-specific
extensions would be a pretty easy solution to the need for an extension

We could also add a self-managed producer registry, as well.

Note that it's likely that any vendor extension would be renumbered
when it was adopted into a standard.  The impact of that would be
much less than the annoyance of vendor/vendor conflicts.

Ron Brender wrote:
> Any value in a DWARF-defined "vendor-specific area" of a DWARF construct 
> is, by its nature, vendor-specific. Which is to say that to properly 
> interpret such a value the consumer must take into account the producer 
> identified in the unit header. Which is to say that there is no such 
> thing as a "conflict" for values in vendor-specific codes (unless the 
> producer is confused/inconsistent).
> That said, since there is no published registry to identify producers it 
> often feels it would be nice if there were a registry of vendor-specific 
> codes that informal cooperation could exploit to minimize re-use of the 
> same codes, so that the identity of the producer could be ignored (at 
> least usually). So far there is no such registry.
> Could/should there be? Good question. There seems to be enough critical 
> mass of DWARF committee participants these days + a formal framework 
> (FSO) that perhaps it would make sense to establish such a registry. 
> After the initial startup, I would expect on-going administration to be 
> relatively low effort.
> Thoughts?
> Ron
> ------------------------------------------------------------------------
> Kees Bakker wrote:
>> Hi,
>> For our compiler and debugger we need a couple of "user" values
>> in dwarf.h
>> We don't want to get into a situation where we will pick a value
>> and later discover that someone else just picked that same value
>> too. (As did happen, I guess, with some MIPS and HP attributes.)
>> Oh, too late, we already did (DW_ATE_Tasking_fract).
>> Is it enough to just report it to the list here? Or is there a
>> more official procedure?
>> Some values that we have been using lately are:
>> DW_TAG_Tasking_circ_type    0x5101
>> DW_TAG_Tasking_mwa_circ_type    0x5102
>> DW_TAG_Tasking_rev_carry_type    0x5103
>> DW_TAG_Tasking_rom        0x5111
>> DW_AT_Tasking_compat_v2        0x2301
>> DW_ATE_Tasking_fract        0x80 (conflicts with HP)
>> DW_ATE_Tasking_accum        0x81 (conflicts with HP)
>> DW_LANG_Tasking_Assembler    0x9101
>> DW_CC_Tasking_interrupt      0x65
>> DW_CC_Tasking_near_system_stack 0x66
>> DW_CC_Tasking_near_user_stack     0x67
>> DW_CC_Tasking_huge_user_stack     0x68
> _______________________________________________
> Dwarf-discuss mailing list
> Dwarf-discuss at lists.freestandards.org
> http://lists.freestandards.org/cgi-bin/mailman/listinfo/dwarf-discuss

Michael Eager	 eager at eagercon.com
1960 Park Blvd., Palo Alto, CA 94306  650-325-8077

More information about the Dwarf-Discuss mailing list