Config::Model::Tester - Test framework for Config::Model


version 4.006


In your test file (typically t/model_test.t):

 use warnings;
 use strict;

 use Config::Model::Tester ;
 use ExtUtils::testlib;

 run_tests() ;

Run tests with:

 perl t/model_test.t [ --log ] [--error] [--trace] [ subtest [ test_case ] ]


This class provides a way to test configuration models with tests files. This class was designed to tests several models and run several tests cases per model.

A specific layout for test files must be followed.

Sub test specification

Each subtest is defined in a file like:


This file specifies that app-name (which is defined in lib/Config/Model/*.d directory) is used for the test cases defined in the * file. The model to test is inferred from the application name to test.

This file contains a list of test case (explained below) and expects a set of files used as test data. The layout of these test data files is explained in next section.

Simple test file layout

Each test case is represented by a configuration file (not a directory) in the *-examples directory. This configuration file is used by the model to test and is copied as $confdir/$conf_file_name using the test data structure explained below.

In the example below, we have 1 app model to test: lcdproc and 2 tests cases. The app name matches the file specified in lib/Config/Model/*.d directory. In this case, the app name matches lib/Config/Model/system.d/lcdproc

 |-- model_test.t
 \-- model_tests.d           # do not change directory name
     |--   # subtest specification for lcdproc app
     \-- lcdproc-examples
         |-- t0              # test case t0
         \-- LCDD-0.5.5      # test case for older LCDproc

Subtest specification is written in file (i.e. this module looks for files named like <app-name>>).

Subtests data is provided in files in directory lcdproc-examples ( i.e. this modules looks for test data in directory <model-name>-examples>. contains instructions so that each file is used as a /etc/LCDd.conf file during each test case. can contain specifications for more test cases. Each test case requires a new file in lcdproc-examples directory.

See "Examples" for a link to the actual LCDproc model tests

Test file layout for multi-file configuration

When a configuration is spread over several files, each test case is provided in a sub-directory. This sub-directory is copied in conf_dir (a test parameter as explained below)

In the example below, the test specification is written in Dpkg layout requires several files per test case. contains instructions so that each directory under dpkg-examples is used.

 \--         # subtest specification
 \-- dpkg-examples
     \-- libversion            # example subdir, used as test case name
         \-- debian            # directory for used by test case
             |-- changelog
             |-- compat
             |-- control
             |-- copyright
             |-- rules
             |-- source
             |   \-- format
             \-- watch

See "Examples" for a link to the (many) Dpkg model tests

More complex file layout

Each test case is a sub-directory on the *-examples directory and contains several files. The destination of the test files may depend on the system (e.g. the OS). For instance, system wide ssh_config is stored in /etc/ssh on Linux, and directly in /etc on MacOS.

These files are copied in a test directory using a setup parameter in test case specification:

  setup => {
    test_file_in_example_dir => 'destination'

Let's consider this example of 2 tests cases for ssh:

 |-- ssh-examples
     \-- basic
         |-- system_ssh_config
         \-- user_ssh_config

Unfortunately, user_ssh_config is a user file, so you need to specify where is located the home directory of the test with another global parameter:

  home_for_test => '/home/joe' ;

For Linux only, the setup parameter is:

 setup => {
   system_ssh_config => '/etc/ssh/ssh_config',
   user_ssh_config   => "~/.ssh/config"

On the other hand, system wide config file is different on MacOS and the test file must be copied in the correct location. When the value of the setup hash is another hash, the key of this other hash is used as to specify the target location for other OS (as returned by Perl $^O variable:

      setup => {
        'system_ssh_config' => {
            'darwin' => '/etc/ssh_config',
            'default' => '/etc/ssh/ssh_config',
        'user_ssh_config' => "~/.ssh/config"

systemd is another beast where configuration files can be symlinks to /dev/null or other files. To emulate this situation, use an array as setup target:

  setup => {
      # test data file => [ link (may be repeated), ..       link(s) target contains test data ]
      'ssh.service' => [ '/etc/systemd/system/sshd.service', '/lib/systemd/system/ssh.service' ]

This will result in a symlink like:

   -> /absolute_path_to/wr_root/model_tests/test-sshd-service/lib/systemd/system/ssh.service

See the actual Ssh and Sshd model tests

Basic test specification

Each model subtest is specified in <app> This file must return a data structure containing the test specifications. Each test data structure contains global parameters (Applied to all tests cases) and test cases parameters (parameters are applied to the test case)

 use strict;
 use warnings;
   # global parameters

   # config file name (used to copy test case into test wr_root/model_tests directory)
   conf_file_name => "fstab",
   # config dir where to copy the file (optional)
   conf_dir => "etc",
   # home directory for this test
   home_for_test => '/home/joe'

   tests =>  [
       # test case 1
       name => 'my_first_test',
       # other test case parameters
       # test case 2
       name => 'my_second_test',
       # other test case parameters
     # ...

 # do not add 1; at the end of the file

In the example below, t0 file is copied in wr_root/model_tests/test-t0/etc/fstab.

 use strict;
 use warnings;
   # list of tests.
   tests => [
       # test name
       name => 't0',
       # add optional specification here for t0 test
       name => 't1',
       # add optional specification here for t1 test

You can suppress warnings by specifying no_warnings => 1 in each test case. On the other hand, you may also want to check for warnings specified to your model. In this case, you should avoid specifying no_warnings here and specify warning tests or warning filters as mentioned below.

See actual fstab test.

Skip a test

A test file can be skipped using skip global test parameter.

In this example, test is skipped when not running on a Debian system:

 eval { require AptPkg::Config; };
 my $skip = ( $@ or not -r '/etc/debian_version' ) ? 1 : 0;

   skip => $skip,
   tests => [ ] ,

Internal tests or backend tests

Some tests require the creation of a configuration class dedicated for test (typically to test corner cases on a backend).

This test class can be created directly in the test specification by specifying tests classes in config_classes global test parameter in an array ref. Each array element is a data structure that use create_config_class parameters. See for instance the layer test or the test for shellvar backend.

In this case, no application exist for such classes so the model to test must be specified in a global test parameter:

  return {
    config_classes => [ { name => "Foo", element => ... } , ... ],
    model_to_test => "Foo",
    tests => [ ... ]

Test specification with arbitrary file names

In some models, like Multistrap, the config file is chosen by the user. In this case, the file name must be specified for each tests case:

   tests => [ {
       name        => 'arm',
       config_file => '/home/foo/my_arm.conf',
       check       => {},

See the actual multistrap test.

Backend argument

Some application like systemd requires a backend argument specified by user (e.g. a service name for systemd). The parameter backend_arg can be specified to emulate this behavior.

Re-use test data

When the input data for test is quite complex (several files), it may be interesting to re-use these data for other test cases. Knowing that test names must be unique, you can re-use test data with data_from parameter. For instance:

  tests => [
        name  => 'some-test',
        # ...
        name  => 'some-other-test',
        data_from  => 'some-test',    # re-use data from test above
        # ...

See plainfile backend test for a real life example.

Test scenario

Each subtest follow a sequence explained below. Each step of this sequence may be altered by adding test case parameters in <model-to-test>

  • Setup test in wr_root/model_tests/<subtest name>/. If your configuration file layout depend on the target system, you will have to specify the path using setup parameter. See "Test file layout depending on system".

  • Create configuration instance, load config data and check its validity. Use load_check => 'no' if your file is not valid.

  • Check for config data warnings. You should pass the list of expected warnings that are emitted through Log::Log4Perl. The array ref is passed as is to the expect function of "expect" in Test::Log::Lo4Perl. E.g:

        log4perl_load_warnings => [
             [ 'Tree.Node', (warn => qr/deprecated/) x 2 ]  ,
             [ 'Tree.Element.Value' , ( warn => qr/skipping/) x 2 ]

    The Log classes are specified in cme/Logging.

    Log levels below "warn" are ignored.

    Note that log tests are disabled when --log option is used, hence all warnings triggered by the tests are shown.

    Config::Model is currently transitioning from traditional "warn" to warn logs. To avoid breaking all tests based on this module, the warnings are emitted through Log::Log4Perl only when $::_use_log4perl_to_warn is set. This hack will be removed once all warnings checks in tests are ported to log4perl checks.

  • DEPRECATED. Check for config data warning. You should pass the list of expected warnings. E.g.

        load_warnings => [ qr/Missing/, (qr/deprecated/) x 3 , ],

    Use an empty array_ref to mask load warnings.

  • Optionally run update command:

        update => {
             [ returns => 'foo' , ]
             no_warnings => [ 0 | 1 ], # default 0
             quiet => [ 0 | 1], # default 0, passed to update method
             loag4perl_update_warnings => [ ... ] # Test::Log::Log4Perl::expect arguments


    • returns is the expected return value (optional).

    • no_warnings to suppress the warnings coming from Config::Model::Value. Note that no_warnings => 1 may be useful for less verbose test.

    • quiet to suppress progress messages during update.

    • log4perl_update_warnings is used to check the warnings produced during update. The argument is passed to expect function of Test::Log::Log4Perl. See load_warnings parameter above for more details.

    • DEPRECATED. update_warnings is an array ref of quoted regexp (See qr operator) to check the warnings produced during update. use update => [] to check that no warnings are issued during update.

    All other arguments are passed to update method.

  • Optionally load configuration data. You should design this config data to suppress any error or warning mentioned above. E.g:

        load => 'binary:seaview Synopsis="multiplatform interface for sequence alignment"',

    See Config::Model::Loader for the syntax of the string accepted by load parameter.

  • Optionally, run a check before running apply_fix (if any). This step is useful to check warning messages:

       check_before_fix => {
          dump_errors   => [ ... ] # optional, see below
          load4perl_dump_warnings => [ ... ] # optional, see below

    Use dump_errors if you expect issues:

      check_before_fix => {
        dump_errors =>  [
            # the issues  and a way to fix the issue using Config::Model::Node::load
            qr/mandatory/ => 'Files:"*" Copyright:0="(c) foobar"',
            qr/mandatory/ => ' License:FOO text="foo bar" ! Files:"*" License short_name="FOO" '

    Likewise, specify any expected warnings:

      check_before_fix => {
            log4perl_dump_warnings => [ ... ],

    log4perl_dump_warnings passes the array ref content to expect function of Test::Log::Log4Perl.

    Both log4perl_dump_warnings and dump_errors can be specified in check_before_fix hash.

  • Optionally, call apply_fixes:

        apply_fix => 1,
  • Call dump_tree to check the validity of the data after optional apply_fix. This step is not optional.

    As with check_before_fix, both dump_errors or dump_warnings can be used.

  • Run specific content check to verify that configuration data was retrieved correctly:

        check => {
            'fs:/proc fs_spec' => "proc",
            'fs:/proc fs_file' => "/proc",
            'fs:/home fs_file' => "/home",

    The keys of the hash points to the value to be checked using the syntax described in "grab" in Config::Model::Role::Grab.

    Multiple check on the same item can be applied with a array ref:

        check => [
            Synopsis => 'fix undefined path_max for st_size zero',
            Description => [ qr/^The downstream/,  qr/yada yada/ ]

    You can run check using different check modes (See "fetch" in Config::Model::Value) by passing a hash ref instead of a scalar :

        check  => {
            'sections:debian packages:0' => { mode => 'layered', value => 'dpkg-dev' },
            'sections:base packages:0'   => { mode => 'layered', value => "gcc-4.2-base' },

    The whole hash content (except "value") is passed to grab and fetch

    A regexp can also be used to check value:

       check => {
          "License text" => qr/gnu/i,

    And specification can nest hash or array style:

       check => {
          "License:0 text" => qr/gnu/i,
          "License:1 text" => [ qr/gnu/i, qr/Stallman/ ],
          "License:2 text" => { mode => 'custom', value => [ qr/gnu/i , qr/Stallman/ ] },
          "License:3 text" => [ qr/General/], { mode => 'custom', value => [ qr/gnu/i , qr/Stallman/ ] },
  • Verify if a hash contains one or more keys (or keys matching a regexp):

     has_key => [
        'sections' => 'debian', # sections must point to a hash element
        'control' => [qw/source binary/],
        'copyright Files' => qr/.c$/,
        'copyright Files' => [qr/\.h$/], qr/\.c$/],
  • Verify that a hash does not have a key (or a key matching a regexp):

     has_not_key => [
        'copyright Files' => qr/.virus$/ # silly, isn't ?
  • Verify annotation extracted from the configuration file comments:

        verify_annotation => {
                'source Build-Depends' => "do NOT add libgtk2-perl to build-deps (see bug #554704)",
                'source Maintainer' => "what a fine\nteam this one is",
  • Write back the config data in wr_root/model_tests/<subtest name>/. Note that write back is forced, so the tested configuration files are written back even if the configuration values were not changed during the test.

    You can skip warning when writing back with the global :

        no_warnings => 1,
  • Check the content of the written files(s) with Test::File::Contents. Tests can be grouped in an array ref:

       file_contents => {
                "/home/foo/my_arm.conf" => "really big string" ,
                "/home/bar/my_arm.conf" => [ "really big string" , "another"], ,
       file_contents_like => {
                "/home/foo/my_arm.conf" => [ qr/should be there/, qr/as well/ ] ,
       file_contents_unlike => {
                "/home/foo/my_arm.conf" => qr/should NOT be there/ ,
  • Check the mode of the written files:

      file_mode => {
         "~/.ssh/ssh_config"     => oct(600), # better than 0600
         "debian/stuff.postinst" => oct(755),

    Only the last four octets of the mode are tested. I.e. the test is done with $file_mode & oct(7777)

    Note: this test is skipped on Windows

  • Check added or removed configuration files. If you expect changes, specify a subref to alter the file list:

        file_check_sub => sub {
            my $list_ref = shift ;
            # file added during tests
            push @$list_ref, "/debian/source/format" ;

    Note that actual and expected file lists are sorted before check, adding a file can be done with push.

  • Copy all config data from wr_root/model_tests/<subtest name>/ to wr_root/model_tests/<subtest name>-w/. This steps is necessary to check that configuration written back has the same content as the original configuration.

  • Create a second configuration instance to read the conf file that was just copied (configuration data is checked.)

  • You can skip the load check if the written file still contain errors (e.g. some errors were ignored and cannot be fixed) with load_check2 => 'no'

  • Optionally load configuration data in the second instance. You should design this config data to suppress any error or warning that occur in the step below. E.g:

        load2 => 'binary:seaview',

    See Config::Model::Loader for the syntax of the string accepted by load2 parameter.

  • Compare data read from original data.

  • Run specific content check on the written config file to verify that configuration data was written and retrieved correctly:

        wr_check => {
            'fs:/proc fs_spec' =>          "proc" ,
            'fs:/proc fs_file' =>          "/proc",
            'fs:/home fs_file' =>          "/home",

    Like the check item explained above, you can run wr_check using different check modes.

Running the test

Run all tests with one of these commands:

 prove -l t/model_test.t :: [ --trace ] [ --log ] [ --error ] [ <model_name> [ <regexp> ]]
 perl -Ilib t/model_test.t  [ --trace ] [ --log ] [ --error ] [ <model_name> [ <regexp> ]]

By default, all tests are run on all models.

You can pass arguments to t/model_test.t:

  • Optional parameters: --trace to get test traces. --error to get stack trace in case of errors, --log to have logs. E.g.

      # run with log and error traces
      prove -lv t/model_test.t :: --error --logl
  • The model name to tests. E.g.:

      # run only fstab tests
      prove -lv t/model_test.t :: fstab
  • A regexp to filter subtest E.g.:

      # run only fstab tests foobar subtest
      prove -lv t/model_test.t :: fstab foobar
      # run only fstab tests foo subtest
      prove -lv t/model_test.t :: fstab '^foo$'


Some of these examples may still use global variables (which is deprecated). Such files may be considered as buggy after Aug 2019. Please warn the author if you find one.


In alphabetical order:

  • Cyrille Bollu

Many thanks for your help.



Dominique Dumont


This software is Copyright (c) 2013-2020 by Dominique Dumont.

This is free software, licensed under:

  The GNU Lesser General Public License, Version 2.1, February 1999



The following websites have more information about this module, and may be of help to you. As always, in addition to those websites please use your favorite search engine to discover more resources.

Bugs / Feature Requests

Please report any bugs or feature requests by email to ddumont at, or through the web interface at You will be automatically notified of any progress on the request by the system.

Source Code

The code is open to the world, and available for you to hack on. Please feel free to browse it and play with it, or whatever. If you want to contribute patches, please send me a diff or prod me to pull from your repository :)

  git clone git://