Re: [Help-bash] errexit and attempting to change readonly functions

[Stephane Chazelas]

2019-09-27 12:12:10 +0200, Andreas Kusalananda Kähäri:
> A function can be made read-only with
> 
>     readonly -f func
> 
> This stops it from being unset or redefined:
> 
>     $ func () { echo hello; }
>     bash: func: readonly function
> 
> However, with the errexit shell option set, I would expect the shell
> to exit when doing this, just like it exits when trying to modify a
> read-only variable under errexit:
> 
>     $ set -e
>     $ func () { echo hello; }
>     $ readonly -f func
>     $ func () { echo hi; }
>     bash: func: readonly function
>     $ # I'm still here
> 
>     $ set -e
>     $ readonly var=42
>     $ var=32
>     bash: var: readonly variable
>     (shell terminates)
> 
> What's the rationale behind this, or is it a bug?  Or did I simply miss
> some text in the manual?
[...]

I'd vote for a bug, though the code looks a bit like it may be
intentional.

$ bash5 -ec '""() { echo x;}; echo x'
bash5: `""': not a valid identifier
x

Also doesn't trigger errexit, except in posix mode (by design
from the code it seems).

"readonly -f" is not POSIX, so there's nothing POSIX can say
about the behaviour when trying to assign a readonly function

It seems as if bash tried to avoid triggering errexit when it
can get away (wrt posix conformance) with not to.

export() { echo x; }

also causes an error (and triggers errexit) in POSIX mode.

Another example is

readonly x
for x in 1 2; do :; done

(POSIX code) That does trigger errexit.

But:

for ((+;;)); do : ; done

(non-POSIX code)

doesn't.

-- 
Stephane