Perhaps too unlikely to be encountered to be worth looking into but
here's a collection of issues encountered with IFS set to \001 or
\177:
n() { (($#)) && printf '<%q>' "$@"; echo; }
# expected observed changed in
IFS=$'\001'
set -- x
n "${@}"#
On 3/4/19 6:00 PM, Grisha Levit wrote:
> Since the changes in [1], there's an issue with CTLNUL leakage in
> ${param=word} when word has null strings, like:
>
> $ unset v; printf '<%q>' ${v= ''}
> <$'\177'>
Thanks for the report. This is still a WIP, keep the test cases coming.
Chet
--
On 3/4/19 10:53 PM, Grisha Levit wrote:
> The changes in [1] to parameter_brace_expand_rhs (for the fix for [2])
> created (exposed?) an issue with ${var=$*} expansion when $* is
> multiple nulls and IFS is unset:
Yes, it was exposed, since the `=' operator takes some special handling
that affects
