Re: Patch: Include platform.h in gnu/gcj/runtime/natStackTrace.cc

[Andrew Haley <aph at redhat dot com>]

Tom Tromey writes:
 > >>>>> "Andrew" == Andrew Haley <aph@redhat.com> writes:
 > 
 > Ranjit> This patch makes "gnu/gcj/runtime/natStackTrace.cc" include
 > Ranjit> "platform.h" so that, at least on Win32, HAVE_BACKTRACE gets
 > Ranjit> defined and the prototype for backtrace( ) gets specified as well.
 > 
 > Andrew> No.  This is a configure variable, and should be set in configure.in.
 > 
 > In this case, backtrace() is provided by libgcj.
 > I was going to approve this patch.
 > 
 > But I see what you mean: we should just AC_DEFINE HAVE_BACKTRACE in
 > configure.in for the appropriate architecture(s).

Precisely.  Life is complicated enough without autoconf variables
being set to be one thing in include/config.h and another elsewhere.

For example: say you read your program source and discover that some
code is conditional on HAVE_BACKTRACE, so you look in include/config.h
and discover that variable isn't set.  However, it *is* set, whatever
config.h says, just somewhere else.  And because you have two
contradictory defintions of HAVE_BACKTRACE the result depends on what
order platform.h and config.h are included.  ARgh!

Andrew.