Re: Request for Obsoletion: DWARF

[Robert Lipe <robertlipe at usa dot net>]

Mark > I know this was true on SCO, but I think RJL decided to switch
Mark > away from DWARF due to the fact that we had lots of bugs there.

jkj> While this is true it does indeed cause problems. Right now the
jkj> most reliable debugging stuff for SCO is still DWARF1, as the OSR5

That statement is not at all true for languages other than C.  The most
reliable debugging and most exercised code in GCC is Dwarf-II.  Yes, it
doesn't work with OpenServer's provided debugger, but that debugger is,
uuuuh, less than ambitious and doesn't with languages other than C very
well anyway.

Jason suggested disabling dwarf1 for C++, where the code "breaks" most
it seems.  My experience dealing with users is that they find the "use
-g for C and -gsomething else for everything else" to be impractical.

jkj> sco5.h that removes a lot of the old cruft I introduced for making
jkj> it both COFf and ELF aware and I find that when I have DWARF2
jkj> enabled, things fall apart.  Admittedly, I usually only test C to
jkj> any degree, not C++.

Then something is wrong in your hacked up version.  DWARF2 has been
the default for OpenServer since late 2000 and has bootstrapped all
available lanaguages and regression tested fine since then.  (I haven't
tested it lately.)  C++ is the language most likely to taunt the Dwarf1
code.

jkj> So plewase consider this a loud but not bellowing complaint to keep
jkj> the code there :-)

The issue is that most of the time the code doesn't *work* and the few
people that understand that code are losing interest in maintaining it.
I leaned on Jason, Jim, and few others for a long time to keep Dwarf-1
on life support, but it's just not worth it...

RJL