UNIX for Advanced & Expert Users

There is just one, its last version is here: Shell Command Language
There are several portability issues between ksh88 and ksh93, and much more with pdksh, which unfortunately some organisations standardize with.

I write and maintain a lot of shell scripts. I do my best to stick to POSIX constructions. When none is convenient enough, I try to use extensions available in both ksh and bash and I explicitly states the shell to use in the shebang (#!/bin/ksh or #!/bin/bash but never #!/bin/sh).

I too use the Shebang line to make a positive statement about the language of the Script. As jlliagre correctly notes, using #!/bin/sh is ambiguous and definitely to be avoided on Sun Solaris systems (where it invokes the old Bourne Shell). On HP-UX it invokes the Posix Shell and on most modern Linux / MACOS systems it invokes bash. Good start.

The Posix folks really need to come up with a name for their "standard" Shell to make it easy for the rest of the world to adopt the standard. Currently it is a joke. The sheer number of posts on this site which propose script which will not run on the poster's Shell is evidence of what might well be a good standard gone wrong.

On every modern Unix system /bin/sh points to either a POSIX compliant shell, or an almost compliant shell (bash --posix) on most Linux systems except Ubuntu, which uses a compliant shell (dash). So POSIX code should run with #!/bin/sh on the compliant systems and in practice on the Linux systems that use bash --posix as well. AFAIK of the modern Unixen only Solaris upto version 10 uses the Bourne shell for /bin/sh. On Solaris 11 this is now ksh93 which is POSIX compliant...

Hi,

Could anyone please explain, how we get to know the ambiguous shebang (#!/bin/sh) invokes which shell in different OS like ubuntu, hp-ux or Solaris?
Thanks!
Kind regards,

It should not matter if you write your script in strict POSIX. All of these shells should then execute your script correctly. Only on Solaris <=10 would you need to use a different location (#!/usr/xpg4/bin/sh).
If you do not use the shebang, then the script will be executed by whatever the calling shell happens to be at the time of execution..


If more functionality than POSIX is required, for example arrays, then one should not use /bin/sh (even if it works), but rather put /bin/bash or /bin/ksh in the shebang or wherever that particular shell happens to reside.

Despite popular belief, the shebang is (currently) to be avoided if you want to write portable POSIX compliant shell scripts. The standard states:

If the first line of a file of shell commands starts with the characters "#!" , the results are unspecified.

You just have to make the script executable and call it from a POSIX environment, i.e. one which has its PATH set to have the POSIX commands first and with some other cleanup, like that one:

Code:

\unalias -a
unset -f command
IFS=`command -p printf " \t\n"`
PATH="`command -p getconf PATH`:$PATH"
export PATH

Unfortunately, this isn't sufficient as you might need extra specific prerequisites like setting a variable telling what version of the standard to follow under HP-UX:

Code:

UNIX_STD=1995; export UNIX_STD

or

Code:

UNIX_STD=2003; export UNIX_STD

It is unfortunate people involved in defining this standard hasn't yet managed to provide a simple and reliable way to define a POSIX script.

I have read a suggestion like that one in the working group discussions:

Code:

#!/bin/SUSv4-POSIX -
PATH=$(getconf PATH)
... posix compliant script starts here

That's kind of a 'warning: water is wet' kind of thing. #! lets you pick awk, perl, or csh, let alone a merely non-POSIX shell.