> On Tue, 13 Mar 2012, David Laight wrote: > >> hmmmm, host sort(1) dependancy. perhaps we need a TOOL_SORT. > > > >Posix deprected the +n terms (even our sort hacks argv[]). > >Shame, because they are what my fingers know about. > > I haven't yet seen a problem report that indicates a need for > a TOOL_SORT. Until now, it has always been enough to use the > POSIXish "-k" options instead of the traditional "+" and "-" > options
yeah, but we're depending upon a host tool that isn't documented. we tend to try to fix those, espcially for the reasons like above. if some host sort(1) has a different output to what we expect then it's not good enough, and there's nothing to stop this happening afaict. eg, we have TOOL_SED and TOOL_AWK now... i'm not saying we _should_ do this, but more asking if others have opinions on it. i'm not sure we should only fix the build issues we see bugs for -- future sort uses might be added that don't break normal builds, but do others, unless we future proof sort usage.. .mrg.