Re: gcc-4.8 and libbacktrace

[Ryan Johnson <ryan dot johnson at cs dot utoronto dot ca>]

On 27/05/2013 11:03 AM, Corinna Vinschen wrote:
> On May 27 10:45, Ryan Johnson wrote:
> > On 27/05/2013 10:22 AM, Corinna Vinschen wrote:
> > > On May 27 10:11, Ryan Johnson wrote:
> > > > Hi all,
> > > >
> > > > FYI, gcc-4.8 includes a new libbacktrace.a (BSD-licensed, I believe)
> > > > that uses dwarf unwind info to accurately unwind the stack.
> > > > Thoughts?
> > > No, since you didn't say what you expect.  What is the end result you're
> > > looking for, a libbacktrace package (Yes, SHTDI), or integration of the
> > > backtrace stuff into Cygwin?  If the latter, that's impossible due to
> > > licensing reasons.
> > The latter, so I guess it's a non-starter.
> >
> > Out of curiosity, what aspect of the library's BSD license allows
> > gcc to use it but not cygwin?
> Oh, BSD.  When I finished reading your email I had already forgotton,
> so I assumed GPLv3+.
>
> So, well, still, SHTDI (not me).  I'm just wondering if it's really
> worth it.  For a normal backtrace we have the stackdump file, which is
> acurate most of the time, for dwarf2-based backtraces you typically run
> the executable under GDB.  And then, you need the debuginfo package as
> well for this kind of debugging on distro packages.
Good point. Come to think of it, the already-available <unwind.h> 
functionality supplied by gcc gives most of the benefit (accurate, 
robust unwinding even when frame pointer was omitted, etc.). That would 
be easier to integrate, too, if it's not already there.

Ryan