Re: Java docs patch: Update libgcj system properties docuementation

[Ranjit Mathew <rmathew at gmail dot com>]

On 4/28/05, Andrew Haley <aph@redhat.com> wrote:
> Ranjit Mathew writes:
>  >
>  > I think it's high time we ditch addr2line and go
>  > with something like Casey Marshall's DWARF-2 reader
> 
> No argument there.  The only thing that stands in the way of doing
> this is testing on a wide range of systems.  This means 32-and 64-bit,
> big- and little- endian, and systems that use stabs by default.

So the plan could be:

1. Casey posts his updated patch along with the glue
to make the stack-trace infrastructure use it as well
as deprecating addr2line. Instead of just a DWARF-2
reader, he might want to use the Factory pattern and
just provide the DWARF-2 reader as the initial provider
with people on other platforms providing stabs, etc.
readers, if needed.

2. Once his basic approach and patch is approved,
he can post a heads-up as well as a Request For
Testing (RFT) and RFC on the GCJ and GCC lists.

3. If there are no serious issues or if the issues
found are suitably addressed, we can integrate
this into GCJ.

I know that the "if" in #3 is a big one, but I don't
see how else we can make progress on this
issue.

For better or for worse, Java libraries and 
programs throw a lot of exceptions and many
a Java programmer almost relies on stack-traces
to help debug his program. Sad and ugly, but
as far as I have seen, quite prevelant.

Using addr2line not just slows down the
program and creates "zombie" processes,
but on non-binutils-based-platforms (Win32),
causes surprising results depending on
whether it happens (or not) to be on the PATH.

My 2p.
Ranjit.
-- 
Ranjit Mathew      Email: rmathew AT gmail DOT com

Bangalore, INDIA.    Web: http://ranjitmathew.hostingzero.com/