David Laight wrote:

> I doubt that the ksh in solaris will match that from the CPL sources,
> it is probably based on some older sources...

Right you are - Grant Beattie checked that last night, and they're still
distributing ksh88, for the love of Pete.  There's another very likely
exlanation for why a modern ksh script fails on Solaris - even pdksh is
closer to modern ksh than ksh88 is.

Of course CPL means Sun could easily include a current ksh now, too.

It turns out there is a shell on Solaris that can run (early) ksh93
scripts: /usr/dt/bin/dtksh, built on ksh93d (not quite ten years old).

ksh93d is only lacking roughly the following features (and a bunch of
fixes) compared to the current ksh:  :)

  atan2, hypot, fmod, and pow added to the floating operations
  current extension API including lib_init()
  tcp connections by host name
  pipefail option
  library-path management
  date/time computations
  manpage/html-generation from getopts
  command substitution working even when fs is readonly
  correct globbing when fs is case-insensitive
  non-alpha auto-field-separators in printf
  += variable assignments
  i18n/l10n support
  html/xml entity reference escaping
  re conversion between ksh and egrep syntax
  minimal as well as maximal re matching
  .sh.match
  .paths
  tcp/udp connections to named services
  max array index increased from 4096 to 16M
  read -N
  .sh.file  .sh.fun  .sh.subshell  .sh.command
  long-double arithmetic where supported
  option globstar
  append disciplines
  BUILTIN_LIB

:)

-Chap