Bash declare built-in regression loses array quotations

Eric Blake
Fri Jan 13 20:11:00 GMT 2017

On 01/13/2017 02:00 PM, Matt D. wrote:
> I updated yesterday to the latest Cygwin x86 release which included an
> updated Bash 4.4.5(1). This release changes the default behavior of the
> built-in 'declare' command which is causing scripts which relied on this
> functionality to break.
> The following test will illustrate the change:
> declare -A list
> list[a]=1
> list[b]=2
> list[c]=3
> declare -p | grep list=
> On Bash 4.4.5(1) the output is:
> declare -A list=([a]="1" [b]="2" [c]="3" )
> Previously the result was:
> declare -A list='([a]="1" [b]="2" [c]="3" )'

This is an intentional upstream behavior change.  Quoting the upstream
release notes:

>> There are a few incompatible changes between bash-4.3 and bash-4.4.  Bash
>> now retains the exit status only of asynchronous jobs, as opposed to all
>> jobs.  This means that it is not possible to use `wait' to retrieve the
>> status of a previously-completed synchronous command.  Bash no longer
>> attempts to perform compound assignment if a variable expansion on the
>> right-hand side of an assignment statement to `declare' or `local' has the
>> form of a compound assignment.  There are other changes to compound array
>> assignments used as arguments to the `declare' and `local' builtins, but
>> those should be backwards compatible, albeit with warning messages about
>> deprecated constructs.  Read the descriptions of the various compatibility
>> options in COMPAT or the man page for more details. 
>> q.  The declare builtin no longer displays array variables using the compound
>>     assignment syntax with quotes; that will generate warnings when re-used as
>>     input, and isn't necessary.

Both output forms should produce the same results when re-evaluated;
except that the form with the extra '' now issues warnings (which is why
declare -p no longer uses the extra '').

> I can confirm that this has always been the expected output when running
> this command on Cygwin and Linux. I run CentOS which does not track the
> latest Bash release and I don't know the procedure to check whether this
> is an upstream regression or not.

It's not a regression, but an intentional change; you'll have to update
your script to quit relying on unspecified output.

Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 604 bytes
Desc: OpenPGP digital signature
URL: <>

More information about the Cygwin mailing list