Using C.UTF-8 in Ubuntu fails

[maglub]

I am really not sure if this is a bug in the Ansible module locale_gen.py, or if it is a bug in this role's configure.yml

Introduction

I am trying to create a database with collate C.UTF-8, which fails in the configure.yml. I am using Ubuntu, but I think this applies to Debian as well. This fails with the error "The locales you've entered is not available on your system.", see below.

- name: PostgreSQL | Ensure the locale for lc_collate and lc_ctype is generated | Debian
  become: yes
  locale_gen: name="{{ item }}" state=present
  with_items:
    - "{{ postgresql_locale }}"
    - "{{ postgresql_ctype }}"
  when: ansible_os_family == "Debian"

Possible fix:

  when: ansible_os_family == "Debian" &&  item != "C.UTF-8"

How to reproduce

To check this on Ubuntu, create a playbook with a construct that mimics configure.yml, you can call it collate_check.yml

---
- hosts: all
  tasks:
    - name: PostgreSQL | Ensure the locale for lc_collate and lc_ctype is generated | Debian
      become: yes
      locale_gen: name="{{ item }}" state=present
      with_items:
        - "{{ postgresql_locale }}"
        - "{{ postgresql_ctype }}"
  vars:
    - postgresql_locale: "C.UTF-8"
    - postgresql_ctype: "C.UTF-8"

• Run it:

$ ansible-playbook -i my_inventory collate_check.yml -l myUbuntuHost
PLAY [all] ************************************************************************************************************************************************************************************************

TASK [Gathering Facts] ************************************************************************************************************************************************************************************
ok: [myUbuntuHost]

TASK [PostgreSQL | Ensure the locale for lc_collate and lc_ctype is generated | Debian] *******************************************************************************************************************
failed: [myUbuntuHost] (item=C.UTF-8) => {"changed": false, "item": "C.UTF-8", "msg": "The locales you've entered is not available on your system."}
failed: [myUbuntuHost] (item=C.UTF-8) => {"changed": false, "item": "C.UTF-8", "msg": "The locales you've entered is not available on your system."}

PLAY RECAP ************************************************************************************************************************************************************************************************
myUbuntuHost         : ok=1    changed=0    unreachable=0    failed=1   

Ansible plugin locale_gen.py

This is the section where it goes wrong in locale_gen.py

def is_available(name, ubuntuMode):
    """Check if the given locale is available on the system. This is done by
    checking either :
    * if the locale is present in /etc/locales.gen
    * or if the locale is present in /usr/share/i18n/SUPPORTED"""
    if ubuntuMode:
        __regexp = r'^(?P<locale>\S+_\S+) (?P<charset>\S+)\s*$'
        __locales_available = '/usr/share/i18n/SUPPORTED'
    else:
        __regexp = r'^#{0,1}\s*(?P<locale>\S+_\S+) (?P<charset>\S+)\s*$'
        __locales_available = '/etc/locale.gen'

    re_compiled = re.compile(__regexp)
    fd = open(__locales_available, 'r')
    for line in fd:
        result = re_compiled.match(line)
        if result and result.group('locale') == name:
            return True
    fd.close()
    return False

The offending statement there is the regex, which is matching for something with an undescrore "_", which C.UTF-8 does not have.

Random thoughts

C.UTF-8 is installed more or less always, no? If it is, we can just ignore that check for C.UTF-8, since it seems to be tricky to get hold of the maintainer of locale_gen.py

In Ubuntu 18.04 LTS (at least) /etc/locale.gen will be created when you reconfigure locales, so it will more or less always be there. But it has been discussed in the Ansible community.

• locale_gen fails to generate a locale on Ubuntu if locale.gen exists ansible/ansible-modules-extras#425
• locale_gen fails to generate a locale on Ubuntu if locale.gen exists ansible/ansible#29200

The bugs above is mainly pointing to that locale_gen.py does not enter "ubuntuMode" if the file /etc/locale.gen exists, which is not really an issue in "our" case. Both regexps above will give us the same problem. It will not find "C.UTF-8", even if it is in the file.

[gclough]

Fixed by #431 .