[patch] make dbus-launch export vars, usage bits

Havoc Pennington hp@redhat.com
Sat, 17 May 2003 22:51:39 -0400

On Sat, May 17, 2003 at 10:42:11PM -0400, Colin Walters wrote: 
> Ok, done.  What do you think about adding a check for a system strdup in
> configure.in, and using it if available?  Probably do this for
> _dbus_strdup as well?  If we wanted to make it a bit more efficient we
> could add some checks for an "inline" keyword and use it if available
> too.

I'm generally for the GLib-style "wrapper" approach
(i.e. _dbus_strdup) rather than the traditional GNU "check for
strdup() and otherwise declare our own function called strdup()"

For whether to chain the wrapper to a system implementation when
available, I think the issue is whether system implementations are
ever significantly better than the fallback. i.e. is system strdup()
faster than _dbus_strdup() - I tend to doubt it, but we could look at
the libc sources and see.

For inline, I'd suggest we do what GLib does here (have a look at its
configure.in checks etc.). I haven't done it since I haven't had much
worth inlining anyhow. (newer gcc seems to automatically inline most
things that should be)