The recent switch to the getopt-gnu module weakened getopt.m4 to no longer require optind=0 nor reject optreset=1 as ways to reset internal state. However, coreutils currently uses optind=0 (in at least env.c), even though I noticed that getopt.m4 and the gnulib unit tests are now careful to define OPTIND_MIN to 1 for BSD systems. Is coreutils' usage pattern going to break on a new enough BSD where the native getopt is now selected (where we previously used to force GNU getopt because native getopt had optreset)?
Is it worth adding a new module getopt-reset and adding a function call getopt_reset() that portably resets getopt internal state (either by optind=0 or optreset=1) when the native getopt implementation is in use and does not support optind=0? -- Eric Blake
