More debug info test directories?

More debug info test directories?

[Andrew Cagney]

Hello,

The gdb.stabs directory, while weird :-^, is probably a good unit test. 
  What do people think of additional debug info directories vis:

gdb.dw2
gdb.dw1

(I'm desperate for better names :-) that contain files of raw debug 
info.  It would let people test the various readers without needing a 
corresponding compiler.

thoughts?
Andrew

Re: More debug info test directories?

[Daniel Jacobowitz]

On Tue, Aug 19, 2003 at 09:24:27AM -0400, Andrew Cagney wrote:
> Hello,
> 
> The gdb.stabs directory, while weird :-^, is probably a good unit test. 
>  What do people think of additional debug info directories vis:
> 
> gdb.dw2
> gdb.dw1
> 
> (I'm desperate for better names :-) that contain files of raw debug 
> info.  It would let people test the various readers without needing a 
> corresponding compiler.
> 
> thoughts?

Absolutely a good idea.  What do you think about letting these use the
.inc files in gdb.asm?  Writing raw dwarf2 could require arch hooks,
unless you want to point all offsets at constant addresses instead of
at a variable's actual location.  For instance some architectures need
to use .word vs .quad.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: More debug info test directories?

[Andrew Cagney]

> Absolutely a good idea.  What do you think about letting these use the
> .inc files in gdb.asm?  Writing raw dwarf2 could require arch hooks,
> unless you want to point all offsets at constant addresses instead of
> at a variable's actual location.  For instance some architectures need
> to use .word vs .quad.

Ah.  If it's needed.

Andrew

Re: More debug info test directories?

[Michael Elizabeth Chastain]

> What do people think of additional debug info directories vis:
> 
> gdb.dw2
> gdb.dw1

Sounds good to me.

Why not call them gdb.dwarf-1 and gdb.dwarf-2?  Oh, because of
the 8.3 uniqueness requirement.  Sigh.

Well, you could re-visit the issue of dwarf-1 support first.  It may be
too early to remove dwarf-1 support, but I think it's okay to not add
any more dwarf-1 tests.  The existing test suite gave, like, 50 ERRORs
and 3500 FAILs the last time I ran it.  And the only dwarf-1 compilers
I know about are Absoft, Diab, and SCO's version of gcc.

Michael C