Re: Incorrect / Inconsistent behavior with nameref assignments in functions

[Binarus]

Thank you very much for your effort, testing and support!

On 29.08.2020 01:46, Koichi Murase wrote:
> 2020-08-28 22:04 Binarus <lists@binarus.de>:
>> Description:
>> ------------
>>
>> Under certain circumstances, assignments of namerefs to local variables
>> in functions behaves in a way which makes namerefs completely useless.
>> Furthermore, the behavior is not consistent.
> 
> This is actually not related to namerefs and has already been fixed in
> Bash 5.1 and the devel branch.  Think about the following codes:
> 
>   a=1; f1() { local a=$a; local; }; f1
>   a=2; f2() { local -a a=("$a"); local; }; f2
> 
> The results for `f1' are the same for all the Bash versions
> 2.0..devel, but the results for `f2' varies in versions.  Here is the
> summary of the results from the different versions of Bash:
> 
> - 2.0..3.0: f1: a=1, f2: a=([0]="1")
> - 3.1:      f2: a=1, f2: a=([0]="")
> - 3.2..4.2: f1: a=1, f2: a=([0]="1")
> - 4.3..5.0: f1: a=1, f2: a=([0]="")
> - 5.1..dev: f1: a=1, f2: a=([0]="1")

This is very interesting. I never have written code like that and
therefore had my problem not before making heavy use of namerefs. So I
have tested my code in three different bash versions which all are buggy ...

I am surprised that a bug of such severity could survive several years.
I don't know when 4.3 came out, but my version of 4.4 is from 2016, and
5.1 is not out yet, so the bug survived at least 4 years (not taking
into account devel or beta versions, which are not an option for most
people).

> I checked the detailed changes.  The behavior of `f2' in 3.1 was
> reported as a bug in the following thread.
> 
>   https://lists.gnu.org/archive/html/bug-bash/2006-05/msg00025.html
> 
> It was fixed in 8b35878f (commit bash-20060504 snapshot).  However,
> the bug seems to be introduced again in 36eb585c (commit bash-20121221
> snapshot).  This regression has been reported at
> 
>   https://savannah.gnu.org/support/index.php?109669
> 
> Finally, it was again fixed in c6c7ae81 (commit bash-20200427
> snapshot).

Thank you very much again for that invaluable information! I am
wondering when debian will include bash 5.1. It looks like debian
testing and debian unstable are on bash 5.0, so it will probably take
several years.

Best regards,

Binarus