Re: [rfa] Dwarf2 testsuite

[Daniel Jacobowitz <drow at false dot org>]

On Fri, Oct 08, 2004 at 03:45:28PM -0400, Michael Chastain wrote:
> Okay, I threw this in my in-queue and I'll reply on Sunday (estimated).

Cool.  Thank you.

> Some random comments:
> 
> . It's okay to have tests that execute on some systems and not others,
>   as long as gnu/hurd and gnu/linux are included.  This is actually
>   covered in section 6, "platforms to support", in
>     http://www.fsf.org/prep/maintain/maintain.html

Hmm, Hurd is ELF also, so that should be fine.

> . Makefile.in needs a copyright notice.  Actually all those little
>   subdirectories shouldn't even need configure/makefile stuff,
>   but I don't wanna get into that right now.

OK.

> . The self-documenting bit is cool, just leave a comment
>   "compile the file and run readelf -w to see what is going on".
>   In fact you might want to strip the comments from this file,
>   because when I go to debug something with a file like this,
>   I don't trust the comments in the file anyways.

I'd prefer to leave the comments now that I've written them, for two
reasons:
 (A) !#(*& it, I spent a lot of effort on those comments!
 (B) they make it easier to make a copy with only an incremental change

> . We definitely need some infrastructure to return what the
>   debug format is, but you can just ignore that and do whatever
>   for now, as long as it doesn't totally blow up when I explicitly
>   test with "runtest /gdb:debug_flags=-gstabs+ -g2".

It won't blow up.  In fact, if GDB is working correctly, it will pass -
if it doesn't, that's a bug.  Mixed stabs and dwarf2 ought to work.

-- 
Daniel Jacobowitz