Eric Blake wrote:
Dan Jacobson uncovered what I believe is a bash bug: http://article.gmane.org/gmane.comp.shells.bash.bugs/8574 See my arguments as to why I think this is a violation of POSIX rules: http://article.gmane.org/gmane.comp.shells.bash.bugs/8575 Although the Shell Substitutions portion of the manual already recommends using `foo` instead of $(foo), it may be worth adding to that section to also document that bash 3.0 (I haven't verified whether bash 3.1 has fixed the bug), zsh (as of 4.2.6), and ksh (5.2.14) all have a bug in parsing command substitutions with an
"pdksh", I presume? David Korn's ksh works fine. Meanwhile, rather than use context sensitive coding techniques, always enclose the case patterns in full parenthesis pairs. I don't know about ancient shells, but it seems to work fine on every shell that accepts $(). Why should we care in autoconf anyway, though? Are we not still constrained to pre-1977 Bourne shell syntax? :) Cheers - Bruce _______________________________________________ Autoconf mailing list Autoconf@xxxxxxx http://lists.gnu.org/mailman/listinfo/autoconf