refactor(completions/*): ensure quoting for compgen -W '${arr[@]}'
#887
Conversation
akinomyoga
force-pushed
the
failglob-2
branch
from
b9b84d1
to
b642a95
Compare
akinomyoga
force-pushed
the
failglob-2
branch
from
b642a95
to
ecc1f34
Compare
There was a problem hiding this comment.
This seems the right thing to do in a sense, but given the uses are mostly in cases like COMPREPLY=($(...)) that just end up splitting everything to the array by whitespace anyway, I feel in practice this doesn't make much of a practical difference at the moment, or at least it doesn't generally protect against whitespace embedded in individual completion items. But perhaps I'm missing something?
Ah, yes. I actually intend to submit a separate PR for the mass change of Another reason is that it is not evident at a glance whether these codes are kept unquoted because they are harmless or just not yet checked. It is noisy for me because I tend to check whether it is safe or not every time I witness such a code. |
akinomyoga
force-pushed
the
failglob-2
branch
from
ecc1f34
to
f5fcc29
Compare
|
Just a ping here about the conflict so it doesn't go unnoticed, no rush. |
|
Thank you! I'll update it. |
With the current implementation, an empty completion would be generated when there is a file whose basename is `minirc.`. In this patch, we filter such files in the pathname-expansion phase by requiring at least one character after `minirc.` in the filenames.
* fix a leaked variable `filedir` (written directly in the global
context)
* fix a problem that filenames containing IFS would be splitted
`$(/bin/ls ...)`.
* fix a problem that the fixed path `/bin/ls' always uses the system
`ls' even when the user wants to replace it with another version for
virtual environments, NixOS, etc.
* fix a problem that elements of tmplist would be added multiple times
with `tmplist+=(${tmplist[@]-} new_elem)`
* fix a problem that word splitting and pathname expansions are
performed for `${tmplist[@]-}.`
* fix a problem that the package names can be splitted by (unusual)
IFS
* fix localvar_inherit for `pkginst_list`
|
I have rebased it. |
Since the removal of elements are performed by "${params[@]/$i/}" in
the above loops, the removed elements remain in the array `params` as
empty elements. Therefore, we need to remove the empty elements by
specifying `-X ''`. This is a regression by
scop#887. These empty
elements were originally automatically dropped by unquoted
${params[@]}.
Reference:
scop@13c28e9#diff-62e0a63a79957c52fe6b79be0a66a987da044839c6bcaeaa725fcd199c4e528fR160
Since the removal of elements are performed by "${params[@]/$i/}" in
the above loops, the removed elements remain in the array `params` as
empty elements. Therefore, we need to remove the empty elements by
specifying `-X ''`. This is a regression by
scop#887. These empty
elements were originally automatically dropped by unquoted
${params[@]}.
Reference:
scop@13c28e9#diff-62e0a63a79957c52fe6b79be0a66a987da044839c6bcaeaa725fcd199c4e528fR160
Since the removal of elements are performed by "${params[@]/$i/}" in
the above loops, the removed elements remain in the array `params` as
empty elements. Therefore, we need to remove the empty elements by
specifying `-X ''`. This is a regression by
scop#887. These empty
elements were originally automatically dropped by unquoted
${params[@]}.
Reference:
scop@13c28e9#diff-62e0a63a79957c52fe6b79be0a66a987da044839c6bcaeaa725fcd199c4e528fR160
Since the removal of elements are performed by "${params[@]/$i/}" in
the above loops, the removed elements remain in the array `params` as
empty elements. Therefore, we need to remove the empty elements by
specifying `-X ''`. This is a regression by
scop#887. These empty
elements were originally automatically dropped by unquoted
${params[@]}.
Reference:
scop@13c28e9#diff-62e0a63a79957c52fe6b79be0a66a987da044839c6bcaeaa725fcd199c4e528fR160
There are also other fixes I noticed while rewriting
compgen -W '${arr[@]}'. The descriptions are found in the commit messages.