> > So this patch was not only not fixing a bug, but merely an optimization, > > but also introducing a bug. > > > > Please leave the introduction of bugs to post 2.60 for now, unless you > > also fix actual bugs. Micro-optimization doesn't count for the latter.
Sorry about that - I guess I had not thought about the full consequences of changing semantics of a core macro, especially this close to a release, when my only goal was a faster configure file. > > But we can do better! We can optimize out _all_ extra forks for nonexistent > shells! I hope you will count this as a bug fix! :-) I'll leave that to Ralf or others to decide. > > > * lib/m4sugar/m4sh.m4 (_AS_PATH_WALK): Optimize nonexistent > > directories, unless optional third argument supplied. > > (AS_UNAME): Don't optimize PATH walk. > > On a second thought, I shouldn't have accepted that patch. > If any caller needs to skip nonexistent directories, he can do so by > insering the command > test -d "$ac_dir" || continue > to the beginning of the second parameter. Fair enough. Go ahead and revert my patch (but your proposed patch is too small - you also need to revert my comment changes). -- Eric Blake
