Re: Strange behaviour when sourcing files with functions

[Chet Ramey]

> In this case, the sourced script doesn't manipulate the positional
> parameters - it doesn't use set or shift - so it does seem like a bug.
> OTOH, I can't reproduce it with bash 2.05b with the current patches,
> so it seems to have been fixed already.  (BTW, wouldn't it make more
> sense to make the positional parameters persist only if they were
> manipulated *and* none were explicitly passed to the sourced script?
> Making them persist also when some were explicitly passed seems less
> useful, and wouldn't have affected backward compatibility.)

It was a compromise.  Scripts run with `.' are supposed to behave as
if the commands were entered interactively or appeared inline in some
shell script.  If a script run with `.' establishes a new set of
positional parameters with `set', it expects them to persist after `.'
completes.  That is backwards compatibility.

There is also the issue of whether or not `.' was executed in a shell
function to complicate matters, since function calls explicitly save
and= restore the positional parameters.

Chet


-- 
``The lyf so short, the craft so long to lerne.'' - Chaucer
( ``Discere est Dolere'' -- chet )
                                                Live...Laugh...Love
Chet Ramey, ITS, CWRU    chet@po.CWRU.Edu    http://cnswww.cns.cwru.edu/~chet/