Re: remote dwarf info using libunwind

[Mike Cvet <mcvet at redhat dot com>]

On Wed, 2006-09-06 at 10:24 -0400, Elena Zannoni wrote:
> Mark Wielaard writes:
>  > Hi,
>  > 
>  > On Wed, 2006-09-06 at 10:18 -0400, Mike Cvet wrote:
>  > > Since the ptrace thread was essentially just a workaround for #2595 to
>  > > address problems with ptrace, I'm not sure if it would be a good idea to
>  > > have more dependencies on this thread if we can avoid it. Hopefully this
>  > > thread will disappear altogether in the future via utrace.
>  > > 
>  > > That being said, I'll take a working solution right now if it means work
>  > > can get done immediately on stacktraces, as long as it'll get fixed up
>  > > properly later.
>  > 
>  > When do we expect to start depending on and using utrace? Is there any
>  > other solution except using the ptrace calls of libunwind to get remote
>  > stacktraces working for now?
> 
> Remember that we will have to keep frysk working on older kernels that
> have the old ptrace (RHEL4, fc5), and so we will not delete that hack
> any time soon.

Well it appears all the callbacks were set up properly to route through
the PtraceThread anyway - so this is a good thing. I just spoke with
Andrew and I'll be checking in Alex's changes this afternoon.

The bugs that Tom has seen and others have observed are still present,
but since the remote unwinding is actually /doing/ somewhat correct,
it'll go in now and others can take a look at it if they like.

- Mike