On Fri, Jun 03, 2022 at 05:26:22AM -0400, Ionen Wolkens wrote:
> On Fri, Jun 03, 2022 at 09:14:05AM +0200, Ulrich Mueller wrote:
[...]
> > brittle. Also, don't use eval because it is evil.
>
> This is static evals, they're just evaluating a flat string and it's
> no different than.. well just running it as-is except it works around
> a noise issue (the 2>/dev/null doesn't register without it).
>
> eval is mostly evil when it's:
> eval "${expanded_variable}=${what_is_this_even}"
>
> Seeing eval "var=\${not_expanded}" as "evil" makes no sense to me,
> it's a harmless warning silencer. I feel this is just overreaction
> to the eval keyword (also in other dev ML post).
>
> To reproduce the warning:
> $ var=$(printf "\0") # var=$(<file-with-null-bytes)
> bash: warning: command substitution: ignored null byte in input
>
> doesn't work: var=$(printf "\0") 2>/dev/null
> works: eval 'var=$(printf "\0")' 2>/dev/null
Was experimenting for what else works just now.
{ var=$(printf "\0"); } 2>/dev/null
I feel like I remember trying this before for qa-sed but don't remember
why I didn't use it (unlike subshells it does keep the var readable).
I still don't think that a flat eval was "evil" but if I have a way
to avoid it I'll use that instead.
--
ionen
signature.asc
Description: PGP signature
