github PR to "Add command to improve robustness"

github PR to "Add command to improve robustness"

[Oliver Kiddle]

PR #101 adds "command" in several command substitutions in
_find_net_interfaces:

  https://github.com/zsh-users/zsh/pull/101/files

I would guess that the cause of the "bug" is function wrappers being
defined which wrap ifconfig and ip to apply colouring to their output.

There are a few random other cases where command is used as a
pre-command modifier in the completions like this. I'm not keen on
applying this selectively in one or two instances. And even less keen
on adding it everywhere. Shouldn't it be _call_program's job to apply
this? Though I've no idea how that can cope with pipelines. Or with
the few cases where we use it to run builtins. I've never been keen on
_call_program's use of eval - if the command comes from a style then
fair enough but for the predefined one it mostly isn't helping anything.
Any ideas?

Or should we just tell them to test something like [[ -t 1 ]] if they
insist on writing wrappers for system utilities that add colour.

Oliver

Re: github PR to "Add command to improve robustness"

[Bart Schaefer]

[-- Attachment #1: Type: text/plain, Size: 573 bytes --]

On Wed, Sep 20, 2023, 10:20 AM Oliver Kiddle <opk@zsh.org> wrote:

>
> I would guess that the cause of the "bug" is function wrappers being
> defined which wrap ifconfig and ip to apply colouring to their output.
>
> Any ideas?
>

If _call_program runs in a subshell, it could start with

unfunction -m '*'

That might be a bit violent, though.  Maybe a zstyle listing functions to
ignore?

Or should we just tell them to test something like [[ -t 1 ]] if they
> insist on writing wrappers for system utilities that add colour.
>

That would certainly be best practice.

>

[-- Attachment #2: Type: text/html, Size: 1413 bytes --]