Re: 01-as-require-shell-fn.patch

[Eric Sunshine]

On Sat, 13 Dec 2003 09:31:07 +0100, Paolo Bonzini wrote:
> Eric Sunshine wrote:
> > Specifically, more recent Bourne shells localize the variables
> > $*, $@, $1, $2, ..., during a function call, whereas older shells
> > do not.  This means that it is safe to "shift" arguments within a
> > function in more recent shells without affecting outer bindings
> > of these variables, but it is not safe to do so with older
> > shells.
> What if you do not "shift" arguments?  Are $*, $@, $<n> then properly
> restored.

No, they are not. The $*, $@, $<n> variables are clobbered by any and all  
function calls since they are not localized to the function.  For example:

foo() {
    echo "Inner: $*"
}

echo "Outer: $*"
foo "bar" "baz" "cow"
echo "Outer: $*"

When this script is invoked with an older shell, such as the one shipped  
with NextStep, the call to foo() destructively changes the outer bindings of  
$*, etc., as shown below:

% sh ./shell.sh wiffles eat smoo
Outer: wiffles eat smoo
Inner: bar baz cow
Outer: bar baz cow

On the other hand, when invoked with a more recent shell, such as a modern  
version of Bash, the outer bindings of $*, etc., remain intact:

% bash ./shell.sh wiffles eat smoo
Outer: wiffles eat smoo
Inner: bar baz cow
Outer: wiffles eat smoo

-- ES