This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
RE: Constructor order help (related to PR 6482)
- From: "Kean Johnston" <jkj at sco dot com>
- To: "'Richard Henderson'" <rth at redhat dot com>
- Cc: <gcc at gcc dot gnu dot org>
- Date: Fri, 11 Apr 2003 17:58:53 -0700
- Subject: RE: Constructor order help (related to PR 6482)
- Organization: The SCO Group
> Eh. My guess is that the common case is actually that hosts need
> to choose based on some autoconf test, and *that* is easier to do
> in config.gcc.
In that case, shouldn't the machine cases in config.gcc simply
specify the file directly? CRTBE_FILE="crtbe-array.c" for example.
I have always hated the idea of a C file including another. It
always messes the debugger up ... at least so far I have never
seen a debugger set it right. It recognizes that the line numbers
are different but not that the file has changed. Maybe I'll have
more luck now that OSR5 uses DWARF-2. Who knows. But if we can
avoid it we should. At a very early age I was taught that the
golden rule of programming is "code for your debugger". IE if its
not easy to see whats happening from within your debugger, the
code is wrong, even if it has no logic bugs. So far that advice
has panned out. Its been 20-odd years and it hasn't let me down :)
> For any host that really needs preprocessor help, we can create
> a special config/foo/crtbe-bar.c that has just their specific
> hackery.
Indeed, and have their machine entry point to it. If, for those
funky targets they decide they then eed to #include one of the
real crtbe* files, they can do it and good luck to them when
they debug.
Kean