Re: backtrace() vs. _Unwind_Backtrace()

[Andrew Haley <aph at redhat dot com>]

David Daney writes:
 > Andrew Haley wrote:
 > 
 > > > 
 > > > But in that case, HAVE_BACKTRACE should be turned on for ia64 Linux.
 > >
 > >Good.
 > >
 > > > Older versions of libc have a broken backtrace() on that platform,
 > > > but even that can be worked around by doing:
 > > > 
 > > > 	LD_PRELOAD=/usr/lib/libunwind-ia64.so.1
 > > > 
 > > > and with newer versions of libc, backtrace() will use _Unwind_Backtrace().
 > >
 > >But we need to build a version of libgcj for ia64 Linux that works
 > >without needing a special LD_PRELOAD.

 > The approach for the MIPS port of libgcj is to include a backtrace() 
 > replacement that calls _Unwind_Backtrace().  This works because 
 > backtrace() is a weak symbol in glibc.  It sounds like this is 
 > essentially what you are suggesting.

If libc's backtrace starts to work on all systems we can deprecate
calling _Unwind_Backtrace() directly.  But in the meantime, with
broken backtrace() on several targets, we have to do whatever
workarounds are necessary.

The nice thing about _Unwind_Backtrace() is that it works on a wide
range of targets.  Unfortunately, it doesn't work on all targets, so
we'll need to maintain a mechanism for using either backtrace() or
_Unwind_Backtrace().

Given that we have to maintain such a mechanism, we should default to
the fastest working backtrace on each system.

In the case of the ia64, the fastest working backtrace is
_Unwind_Backtrace().

Andrew.