The Perl Toolchain Summit needs more sponsors. If your company depends on Perl, please support this very important event.


App::Yath - Yet Another Test Harness (Test2-Harness) Command Line Interface (CLI)


This is the primary documentation for yath, App::Yath, Test2::Harness.

The canonical source of up-to-date command options are the help output when using $ yath help and $ yath help COMMAND.

This document is mainly an overview of yath usage and common recipes.

App::Yath is an alternative to App::Prove, and Test2::Harness is an alternative to Test::Harness. It is not designed to replace Test::Harness/prove. Test2::Harness is designed to take full advantage of the rich data Test2 can provide. Test2::Harness is also able to use non-core modules and provide more functionality than prove can achieve with its restrictions.


Test2::Harness/App::Yath is is focused on unix-like platforms. Most development happens on linux, but bsd, macos, etc should work fine as well.

Patches are welcome for any/all platforms, but the primary author (Chad 'Exodist' Granum) does not directly develop against non-unix platforms.


Currently windows is not supported, and it is known that the package will not install on windows. Patches are be welcome, and it would be great if someone wanted to take on the windows-support role, but it is not a primary goal for the project.


To use Test2::Harness, you use the yath command. Yath will find the tests (or use the ones you specify) and run them. As it runs, it will output diagnostic information such as failures. At the end, yath will print a summary of the test run.

yath can be thought of as a more powerful alternative to prove (Test::Harness)


These are common recipes for using yath.


    $ yath

Simply running yath with no arguments means "Run all tests for the current project". Yath will look for tests in ./t, ./t2, and ./ and run any which are found.

Normally this implies the test command but will instead imply the run command if a persistent test runner is detected.


Yath has the ability to preload modules. Yath normally forks to start new tests, so preloading can reduce the time spent loading modules over and over in each test.

Note that some tests may depend on certain modules not being loaded. In these cases you can add the # HARNESS-NO-PRELOAD directive to the top of the test files that cannot use preload.


Any module can be preloaded:

    $ yath -PMoose

You can preload as many modules as you want:

    $ yath -PList::Util -PScalar::Util


If your preload is a subclass of Test2::Harness::Runner::Preload then more complex preload behavior is possible. See those docs for more info.



You can turn on logging with a flag. The filename of the log will be printed at the end.

    $ yath -L
    Wrote log file: test-logs/2017-09-12~22:44:34~1505281474~25709.jsonl

The event log can be quite large. It can be compressed with bzip2.

    $ yath -B
    Wrote log file: test-logs/2017-09-12~22:44:34~1505281474~25709.jsonl.bz2

gzip compression is also supported.

    $ yath -G
    Wrote log file: test-logs/2017-09-12~22:44:34~1505281474~25709.jsonl.gz

-B and -G both imply -L.


You can replay a test run from a log file:

    $ yath test-logs/2017-09-12~22:44:34~1505281474~25709.jsonl.bz2

This will be significantly faster than the initial run as no tests are actually being executed. All events are simply read from the log, and processed by the harness.

You can change display options and limit rendering/processing to specific test jobs from the run:

    $ yath test-logs/2017-09-12~22:44:34~1505281474~25709.jsonl.bz2 -v [TEST UUID(S)]

Note: This is done using the $ yath replay ... command. The replay command is implied if the first argument is a log file.


The -T option will cause each test file to report how long it took to run.

    $ yath -T

    ( PASSED )  job  1    t/yath_script.t
    (  TIME  )  job  1    Startup: 0.07692s | Events: 0.01170s | Cleanup: 0.00190s | Total: 0.09052s


yath supports starting a yath session that waits for tests to run. This is very useful when combined with preload.


This starts the server. Many options available to the 'test' command will work here but not all. See $ yath help start for more info.

    $ yath start


This will run tests using the persistent runner. By default, it will search for tests just like the 'test' command. Many options available to the test command will work for this as well. See $ yath help run for more details.

    $ yath run


Stopping a persistent runner is easy.

    $ yath stop


The which command will tell you which persistent runner will be used. Yath searches for the persistent runner in the current directory, then searches in parent directories until it either hits the root directory, or finds the persistent runner tracking file.

    $ yath which

The watch command will tail the runner's log files.

    $ yath watch


You can use preloads with the yath start command. In this case, yath will track all the modules pulled in during preload. If any of them change, the server will reload itself to bring in the changes. Further, modified modules will be blacklisted so that they are not preloaded on subsequent reloads. This behavior is useful if you are actively working on a module that is normally preloaded.


    $ yath init

The above command will create is automatically run by most build utils, in which case only the exit value matters. The generated will run yath and execute all tests in the ./t and/or ./t2 directories. Tests in ./t will ALSO be run by prove but tests in ./t2 will only be run by yath.


You can write a .yath.rc file. The file format is very simple. Create a [COMMAND] section to start the configuration for a command and then provide any options normally allowed by it. When yath is run inside your project, it will use the config specified in the rc file, unless overridden by command line options.

Note: You can also add pre-command options by placing them at the top of your config file BEFORE any [cmd] markers.

Comments start with a semi-colon.

Example .yath.rc:

    -pFoo ; Load the 'foo' plugin before dealing with commands.

    -B ;Always write a bzip2-compressed log

    -PMoose ;Always preload Moose with a persistent runner

This file is normally committed into the project's repo.


Sometimes you want to specify files relative to the .yath.rc so that the config option works from any subdirectory of the project. Other times you may wish to use a shell expansion. Sometimes you want both!

    -I rel(path/to/extra_lib)

This will take the path to .yath.rc and prefix it to the path inside rel(...). If for example you have /project/.yath.rc then the path would become /project/path/to/extra_lib.

    --default-search glob(subprojects/*/t)

This will add a --default-search $_ for every item found in the glob. This uses the perl builtin function glob() under the hood.

    --default-search relglob(subprojects/*/t)

Same as glob() except paths are relative to the .yath.rc file.


You can add a .yath.user.rc file. Format is the same as the regular .yath.rc file. This file will be read in addition to the regular config file. Directives in this file will come AFTER the directives in the primary config so it may be used to override config.

This file should not normally be committed to the project repo.


yath will recognise a number of directive comments placed near the top of test files. These directives should be placed after the #! line but before any real code.

Real code is defined as any line that does not start with use, require, BEGIN, package, or #

good example 1

good example 2
    use strict;
    use warnings;


bad example 1

    # blah


bad example 2

    print "hi\n";





Use this if your test will fail when modules are preloaded. This will tell yath to start a new perl process to run the script instead of forking with preloaded modules.

Currently this implies HARNESS-NO-FORK, but that may not always be the case.



Use this if your test file cannot run in a forked process, but instead must be run directly with a new perl process.



yath usually uses the Test2::Formatter::Stream formatter instead of TAP. Some tests depend on using a TAP formatter. This option will make yath use Test2::Formatter::TAP or Test::Builder::Formatter.


yath can be configured to use the Test2::Plugin::IOEvents plugin. This plugin replaces STDERR and STDOUT in your test with tied handles that fire off proper Test2::Event's when they are printed to. Most of the time this is not an issue, but any fancy tests or modules which do anything with STDERR or STDOUT other than print may have really messy errors.

Note: This plugin is disabled by default, so you only need this directive if you enable it globally but need to turn it back off for select tests.


yath will usually kill a test if no events occur within a timeout (default 60 seconds). You can add this directive to tests that are expected to trip the timeout, but should be allowed to continue.

NOTE: you usually are doing the wrong thing if you need to set this. See: HARNESS-TIMEOUT-EVENT.


yath can be told to alter the default event timeout from 60 seconds to another value. This is the recommended alternative to HARNESS-NO-TIMEOUT


yath can be told to alter the default POSTEXIT timeout from 15 seconds to another value.

Sometimes a test will fork producing output in the child while the parent is allowed to exit. In these cases we cannot rely on the original process exit to tell us when a test is complete. In cases where we have an exit, and partial output (assertions with no final plan, or a plan that has not been completed) we wait for a timeout period to see if any additional events come into


This lets you tell yath that the test file is long-running. This is primarily used when concurrency is turned on in order to run longer tests earlier, and concurrently with shorter ones. There is also a yath option to skip all long tests.

This duration is set automatically if HARNESS-NO-TIMEOUT is set.


This lets you tell yath that the test is medium.

This is the default duration.


This lets you tell yath That the test is short.


This lets you tell yath that the test cannot be run concurrently with other tests. Yath will hold off and run these tests one at a time after all other tests.


This lets you tell yath that the test cannot be run concurrently with other tests of this class. This is helpful when you have multiple tests which would otherwise have to be run sequentially at the end of the run.

Yath prioritizes running these tests above HARNESS-CATEGORY-LONG.


This is the default category.


This lets you tell yath that no other test of type XXX can be run at the same time as this one. You are able to set multiple conflict types and yath will honor them.

XXX can be replaced with any type of your choosing.

NOTE: This directive does not alter the category of your test. You are free to mark the test with LONG or MEDIUM in addition to this marker.



Specify a range of job slots needed for the test to run. If set to a single value then the test will only run if it can have the specified number of slots. If given a range the test will require at least the lower number of slots, and use up to the maximum number of slots.

Example with multiple lines.
    # DASH and space are split the same way.

Or on a single line.



This lets you specify a number (minimum n=1) of retries on test failure for a specific test. HARNESS-RETRY-1 means a failing test will be run twice and is equivalent to HARNESS-RETRY.


Use this to avoid this test being retried regardless of your retry settings.


This section documents the App::Yath module itself.


In practice you should never need to write your own yath script, or construct an App::Yath instance, or even access themain instance when yath is running. However some aspects of doing so are documented here for completeness.

A minimum yath script looks like this:

    BEGIN {
        package App::Yath:Script;

        require Time::HiRes;
        require App::Yath;
        require Test2::Harness::Settings;

        my $settings = Test2::Harness::Settings->new(
            harness => {
                orig_argv       => [@ARGV],
                orig_inc        => [@INC],
                script          => __FILE__,
                start           => Time::HiRes::time(),
                version         => $App::Yath::VERSION,

        my $app = App::Yath->new(
            argv    => \@ARGV,
            config  => {},
            settings => $settings,



It is important that most logic live in a BEGIN block. This is so that goto::file can be used post-fork to execute a test script.

The actual yath script is significantly more complicated with the following behaviors:

pre-process essential arguments such as -D and no-scan-plugins
re-exec with a different yath script if in developer mode and a local copy is found
Parse the yath-rc config files
gather and store essential startup information


App::Yath does not provide many methods to use externally.


This tells App::Yath to generate a subroutine at the specified symbol name which can be run and be expected to return an exit value.

$lib_path = $app->app_path()

Get the include directory App::Yath was loaded from.


The source code repository for Test2-Harness can be found at


Chad Granum <>


Chad Granum <>


Copyright 2020 Chad Granum <>.

This program is free software; you can redistribute it and/or modify it under the same terms as Perl itself.