2009-10-28 09:46:49 -0400, Chet Ramey:
[...]
> >> The Posix grammar has never allowed it (a `function_body' must be a
> >> compound command, and that's what bash implements), and there has
> >> never been sufficient demand to add it as an extension.
> > [...]
> >
> > It's never allowed it to POSIX _scripts_, but it won't make a
> > shell non-conformant to support it (and ksh, pdksh, ash, zsh
> > even posh all do support it), just like most POSIX conformant
> > shells support arrays for instance even though that's not a
> > POSIX feature.
>
> That's why I said it would be added as an extension.
Sorry, I overlooked that bit.
> > bash not supporting it helps for writing POSIX compliant scripts
> > (as it will return an error if one tries to use that
> > non-standard feature), but as bash is the only POSIX shell that
> > doesn't support it, it feels a bit silly.
>
> "Silly" it may be, but there's not really the demand for it.
I can understand it. I was more curious about the origins. After
all, that breaks Bourne backward compatibility (in a shell
called Bourne-again shell)) and it seems more effort to exclude
one form of command rather than include any form of command.
f() for i in 1; do echo test; done
f() { echo test; } > xx
are allowed (and I doubt there has been a lot of demand for
that), and can be confusing all the same (the second one for
instance is not interpreted POSIXly by zsh).
> > I wonder what's the rationale for POSIX not supporting it (is it
> > because of bash?).
>
> That would be the tail wagging the dog, wouldn't it? It's never been
> in the grammar.
OK, but again, why exclude simple commands here? Would there be
some implication? Has there be historical versions of sh that
didn't support it?
I guess I should ask the question on the POSIX list.
Cheers,
Stephane
