Re: Which version of gdb supports gcc 3.0 ABI?

[Daniel Berlin <dberlin at redhat dot com>]

Benjamin Kosnik <bkoz@redhat.com> writes:

> If all you need is symbol demangling use the current 'src' module on
> sources.redhat.com.
> 
> The combination of gdb and g++ has been broken for any serious C++
> debugging for quite some time now.

Yup.
>  I'm not quite sure it's all gdb's
> fault. For months I've tried to explain to Daniel that seeing const
> static data members in template base classes is a basic requirement,
> not an advanced feature.

And I agree, but I haven't had the time to track it down.  I seem to
be one of the only ones who cares about C++ support in GDB enough to
work on it, and this leaves me way too many things to do, and not
enough time to do them. I prioritize based on what is most important,
and right now, that is supporting the v3 abi.

>  Basically, gdb and g++ don't play nicely
> together WRT static data: the end result is that any/all static data
> is invisible or 0x when you look at code through gdb's eyes. Looking
> at objects that have inheritance invariably leads to "virtual base
> class botch" messages in gdb.

I don't see virtual baseclass botch anymore, or at least, i haven't gotten one in the new-abi stuff
i've been working on. 
old-abi has been broken since gcc 2.95 or slightly before, in various
ways.
.
>  Printing member functions from even
> simple classes is hopeless. 
> Trying to probe nested objects is an
> exercise in futility.
> 
Yup.

> People say patches will get in "real soon now" but they never gets
> checked in. Thus, the public vent.

I can't check in the patches until they are finished, because gdb
won't compile, because of how much crap was where it didn't belong.

I also gave up trying to work on other non-C++ parts of gdb because
approval takes forever. It completely removes all incentive i have to
improve gdb when i can't get stuff committed, and have to spend months
trying to get someone to approve it. I've thrown away non-C++ patches
i worked on (like the dwarf2 stuff), because it's just not worth the
aggravation of trying to get someone to look at it.
But, once again,  feel free to help me with C++ support in GDB.
It's going to take me 2 or 3 *years* to overhaul C++ support in gdb to
the point it should be at, because it's not a major consideration, and
hasn't been.
It's like thread debugging. Nobody seems to think it's *very very very
very very* important, except all the users.

I've actually started rewriting gdb in C++, and whenever i get very
frustrated at the way gdb is now, i work on it some more.

>
> use 'printf' and 'gcc -s -O0'
> 
> :(
> 
> I could see some (not all) data objects around 2001-01-15, but since
> then it's been an exercise in frustration.
> 
> -benjamin