bug-bash
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: parser bug in Q_ARITH


From: Chet Ramey
Subject: Re: parser bug in Q_ARITH
Date: Mon, 14 Jul 2025 12:10:22 -0400
User-agent: Mozilla Thunderbird

On 7/9/25 4:57 PM, Robert Elz wrote:

jacaranda$ echo "$BASH_VERSION"
5.3.0(1)-release
jacaranda$ unset hash
jacaranda$ declare -A hash; key=\'\]
jacaranda$ let "hash[\$key]++"; declare -p hash
declare -A hash=(["']"]="1" )

So it is still different, just different different.

`let' is a builtin, so it has a well-defined word expansion sequence. You
should be able to replace it with a shell function and have the same thing
happen as when you call the builtin directly. That exempts it from the
internal quoting the shell can do with ((, since that's a compound command.

So, since you don't have assoc_expand_once set -- what started this entire
discussion back in 2021 in the first place -- the subscript is expanded
twice. `let' sees "hash[$key]++", and the array expansion in the arithmetic
evaluation evaluates the subscript again. You have to quote your way out
of trouble.

This was essentially the entire substance of the discussion in 2021 that
prompted these changes.

And for this, the (( )) version is clearly correct, the '\' that
is inserted before the $ in a string which is (as Chet said) treated
as double quoted, prevents the '$key' from being a variable reference,
and instead makes it be just the 4 characters '$' 'k' 'e' 'y' (and in
which "declare -p" needs to similarly escape the '$' for correctness).

Correct.

--
``The lyf so short, the craft so long to lerne.'' - Chaucer
                 ``Ars longa, vita brevis'' - Hippocrates
Chet Ramey, UTech, CWRU    chet@case.edu    http://tiswww.cwru.edu/~chet/

Attachment: OpenPGP_signature.asc
Description: OpenPGP digital signature


reply via email to

[Prev in Thread] Current Thread [Next in Thread]