bash: fix shellcheck errors

[rsteube]

Fixed shellcheck errors according to documentation:

https://github.com/koalaman/shellcheck/wiki/SC2207
https://github.com/koalaman/shellcheck/wiki/SC2164

untested - tests are ok but someone should verify this is correct

[PapaCharlie]

mapfile is not present in all bash versions. Namely, the out-of-the-box bash for MacOS doesn't have it (it's at a 3.0 version, it looks like mapfile was introduced in 4.0).

read -a COMPREPLY looks like it'll do the trick

[rsteube]

I think i need a project where i can test this. Disabling escape characters using -r (shellcheck warning) and the IFS seems to be required as well for it to work. So it's probably one of:

# For bash 3.x+
array=()
while IFS='' read -r line; do array+=("$line"); done < <(mycommand)

or

# For bash
IFS=" " read -r -a array <<< "$(mycommand)"

Some tests with read -a in a bash shell didn't give me the right result (or i simply used it wrong).

[PapaCharlie]

Yeah it seems read -a really doesn't work as expected. I've tested this on multiple bash versions, and it seems to be the most consistent:

while read -s -d " " c ; do
  COMPREPLY+=("$c")
done < <(compgen -W "${allflags[*]}" -- "$cur")

[rsteube]

Does it need to be cleared first?: COMPREPLY=()

[PapaCharlie]

Bash will automatically create the array variable if you start appending to it:

unset a
a+=("foo")
echo "${a[@]}"
foo

But it wouldn't hurt to make sure it's empty anyway

[rsteube]

Was just worried that it is defined before already, but should be fine then.
Just compiled https://github.com/gohugoio/hugo with it but seems there is still sth. wrong with the completion (https://gist.github.com/rsteube/7647a9a3456cdb102f0b88ba004d74f8).
Currently looking into it.

[rsteube]

while IFS='' read -r c; do works in my shell (5.0.7).
https://gist.github.com/rsteube/6bb3336adf902193997bba048094b870

[rsteube]

3.2.57 looks ok as well: https://hub.docker.com/r/albers/bash-completion-mac

[umarcor]

@spf13, @eparis, can we please have this merged so that Travis CI jobs are fixed for every other PR based on master?

[jharshman]

@rsteube, while I review this, can you squash your commits?

[rsteube]

@jharshman done

[jharshman]

@rsteube thanks, I'm just going to do a bit of testing of the bash completions before I merge this in. I would encourage you to do the same if you haven't yet. I'm not seeing any breaking changes here but just to be sure.

[rsteube]

Recompiled hugo with the squashed version and no changes to https://gist.github.com/rsteube/6bb3336adf902193997bba048094b870 which seemed to be fine in bash 5.0.7 and 3.2.57.

[jharshman]

Validated by recompiling kubectl.