W dniu pon, 05.02.2018 o godzinie 20∶33 -0600, użytkownik R0b0t1
napisał:
> On Sun, Feb 4, 2018 at 10:45 PM, Zac Medico <zmed...@gentoo.org> wrote:
> > On 02/04/2018 07:22 PM, R0b0t1 wrote:
> > > This is everything that shellcheck reported as an error. They are not
> > > as serious as the globbing issue, but it would be a good idea to
> > > change them. They are generally "type" issues (e.g. ">" instead of
> > > "-gt").
> > >
> > > Some changes are shellcheck annotations. Very interesting was:
> > >
> > > eval "$x=(\"\${$x[@]}\" ${QA_PREBUILT//\*/.*})"
> > >
> > > Which looks like a bad array expansion ("$x[@]").
> >
> > I don't see a shellcheck error for that, using shellcheck-0.4.7. Maybe a
> > false positive with an older version?
> >
>
> 0.4.6, I will see if I can check.
>
> > > diff --git a/bin/isolated-functions.sh b/bin/isolated-functions.sh
> > > index b28e44f18..377b32d90 100644
> > > --- a/bin/isolated-functions.sh
> > > +++ b/bin/isolated-functions.sh
> > > @@ -82,7 +82,7 @@ __dump_trace() {
> > > lineno=${BASH_LINENO[${n} - 1]}
> > > # Display function arguments
> > > args=
> > > - if [[ -n "${BASH_ARGV[@]}" ]]; then
> > > + if [[ -n "${BASH_ARGV[*]}" ]]; then
> >
> > I feel like the shellcheck authors might be willing to accept [[ -n
> > ${BASH_ARGV[@]} ]] or [[ ${BASH_ARGV[@]} ]] as correct, since the
> > "Problematic code" that they cite involves an incorrect comparison:
> >
> > https://github.com/koalaman/shellcheck/wiki/SC2199#problematic-code
> >
>
> This example might be more illustrative:
>
> argc () {
> echo $#
> }
>
> argc "${BASH_ARGV[*]}"
> argc "${BASH_ARGV[@]}"
>
>
> Output (./test.sh a b):
>
> 1
> 2
>
>
> Which changes the semantics of the tests in which it is present. It is
> hard to know what [[ is doing; if the same is attempted with [ it
> would be an error for the same reason that globbing produces errors.
> See:
>
> [ "${BASH_ARGV[*]}" ]
> [ "${BASH_ARGV[@]}" ] # Fails; [: b: unary operator expected
> [[ "${BASH_ARGV[*]}" ]]
> [[ "${BASH_ARGV[@]}" ]]
>
>
> It is possible [[ ignores the extra elements. I can't think of a
> reason where this would make the results of the test different. At the
> same time, it might give people the wrong impression of the operation
> of [.
>
> In the sense that [ and [[ represent test(1), anything but the "[*]"
> expansion is incorrect, as the error message indicates. That [[ treats
> its arguments specially because it is a feature of the syntax just
> makes the situation more confusing.
>
They don't. '[' works as a external command whose arguments are subject
to word splitting. '[[' works as a builtin that does word splitting
before expanding variables.
--
Best regards,
Michał Górny
