• -a automated testing - done: automated_testing
• -C skip Changes file (done: skip_changes)
• -d extra debugging stuff (done: debug)
• -D don't build dist (done: skip_dist)
• -h help (won't do)
• -j set parallel test (not yet done, but yes, #38)
• -k skip kwalitee (done) - done: skip_kwalitee
• -m skip MANIFEST check - done: skip_manifest
• -p skip prereqs test - done: skip_prereqs
• -t make and test only
• -T don't run tests (done: skip_tests)
• -v print version and exit (won't do)
• -V skip version check (done: skip_cpan_version_check)

Nice to see progress!

1. I've put the version stuff inside an eval in the dist branch of the else, as I found situations it would crash the debug line should be invoked only once there
2. I added a counter in the progress line. Handy when testing with close to 500 perl versions

--- release     2021-02-26 13:12:25.558026859 +0100
+++ release-lx09        2021-02-26 13:56:27.431257980 +0100
@@ -509,22 +509,20 @@ if ($skip_dist) {
        }
 else {
        $release->dist unless( $release->local_file and -e $release->local_file );
-       $Version = $release->dist_version;
-       $release->_debug( "dist version is <$Version>\n" );
+       $Version = eval { $release->dist_version; } || "?";
        }

-$Version = $release->dist_version;
-
 $release->_debug( "dist version is <$Version>\n" );

-
 # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # #
 # test with a bunch of perls
 unless( $opts{T} ) {
        my $old_perl = $release->get_perl;

-       PERL: foreach my $perl ( $release->perls ) {
-               $release->_print("============Testing with $perl\n");
+       my @perls = $release->perls;
+       my ($n, $N) = (1, scalar @perls);
+       PERL: foreach my $perl ( @perls ) {
+               $release->_print("============ Testing with $perl (", $n++, "/$N)\n");
                $release->set_perl( $perl ) or next;

                $release->clean;

Use something much lighter?

with close to 200 available perl versions to test a module against, my .releaserc looks somewhat like this:

--8<---
cpan_user HMBRAND
automated_testing 1
skip_kwalitee 1
skip_manifest 1
skip_prereqs 1
skip_dist 1
allow_glob_in_perls 1
perls /pro/bin/perl:/usr/bin/perl
:/media/Tux/perls/bin/perl5.6.[12]
:/media/Tux/perls/bin/perl5.8.
:/media/Tux/perls/bin/perl5.1[02468]
:/media/Tux/perls/bin/perl5.2[02468]
:/media/Tux/perls/bin/perl5.21.
:/media/Tux/perls/bin/tperl5.6.2
:/media/Tux/perls/bin/tperl5.8.
:/media/Tux/perls/bin/tperl5.1[02468]
:/media/Tux/perls/bin/tperl5.2[02468]
:/media/Tux/perls/bin/tperl5.21.
:/media/Tux/perls/bin/cperl5.2[01].*
-->8---

If a module depends on something else, like DBI, there is currently no way to tell Module::Release to skip a perl if that install does not have that module installed.

Would it be possible to have a line like

requires DBI,Data::Peek

and have Module::Release skip that perl if those modules are not installed?

Prompt with the differences and elect to continue by committing the new MANIFEST or to bail out.

Joel Berger sent this:

I'm hacking a private method seemingly just to override an SSL value: https://metacpan.org/source/BDFOY/Module-Release-2.12/lib/Module/Release/WebUpload/Mojo.pm#L88

Maybe I should do this: https://metacpan.org/pod/distribution/IO-Socket-SSL/lib/IO/Socket/SSL.pod#IO::Socket::SSL::set_args_filter_hack-code-use_defaults

Right now it assumes two decimal places, changing 1.001 to 1.01

Make a switch to decide if that's dev or user releases, then use PPI::Version to do the dirty work.

If working on a project where new tests are work in progress, and these tests are not (yet) in git nor in the MANIFEST, I do not see this as e reason for release -t to fail. So untracked files IMHO should not cause a FAIL.

=item Check that source control is up-to-date

If there are modified files, added files, or extra files so that
source control complains, fail.

I'd like to see an option (which I would default to be enabled in my .releaserc files), to make that

=item Check that source control is up-to-date

If there are modified files so that source control complains, fail.

I've done that by adding -uno:

sub check_vcs {
	my $self = shift;

	$self->_print( "Checking state of Git... " );

	my $git_status = $self->run('git status -s -uno 2>&1');

But if you think this could help others too, it should be an option

It looks like I don't encode standard input, but I do encode it on the way out. That ends up as mojibake.

Parse the source control log and extract all users, then put those in CONTRIBUTORS and commit the result.

Show that so it's easier to note their efforts in Changes

Currently testing Text::CSV_XS with 145+ versions of perl takes more than an hour.

Most of the consumed time is running the tests (52000+ tests taking 15-20 seconds)

I can see two approaches for a speedup.

1: if the t/ folder is parallel-safe, use -jN

$ prove -wbQ t
All tests successful.
Files=29, Tests=52231, 17 wallclock secs ( 3.12 usr  0.20 sys + 16.47 cusr  0.47 csys = 20.26 CPU)
Result: PASS

$ prove -wbQ -j8 t
All tests successful.
Files=29, Tests=52231, 12 wallclock secs ( 3.67 usr  0.22 sys + 18.01 cusr  0.67 csys = 22.57 CPU)
Result: PASS

$ make test
:
All tests successful.
Files=29, Tests=52231, 17 wallclock secs ( 3.20 usr  0.23 sys + 16.16 cusr  0.42 csys = 20.01 CPU)
Result: PASS

$ env HARNESS_OPTIONS=j8 make test
:
All tests successful.
Files=29, Tests=52231, 12 wallclock secs ( 3.77 usr  0.25 sys + 18.70 cusr  0.50 csys = 23.22 CPU)
Result: PASS

Implementing this is easy:

--- release     2021-03-11 08:56:52.000000000 +0100
+++ release-2   2021-03-24 11:19:51.325116916 +0100
@@ -35,6 +35,10 @@ release - give your Perl distros to the
        # set $ENV{AUTOMATED_TESTING} to a true value
        release -a

+       # Enable parallel testing by setting $ENV{HARNESS_OPTIONS}
+       release -j
+       release -j8
+
 =head1 DESCRIPTION

 This is the prototype program for using C<Module::Release>. You should
@@ -321,8 +325,18 @@ my @extra_opts = split /\s+/, $ENV{RELEA
 push @ARGV, @extra_opts;

 my %opts;
-getopts('aChdptvVkT', \%opts) or $opts{h} = 1;
+getopts('aChdptvVkTj:', \%opts) or $opts{h} = 1;

+if( exists $opts{j} ) {
+       unless( defined $opts{j} ) {
+               delete $opts{h}; # getopts does not support optional arguments like Getopt::Long does in "j|parallel:9"
+               $opts{j} = eval {
+                       require System::Info;
+                       System::Info->new->get_core_count;
+                       } || 9;
+       }
+       $ENV{HARNESS_OPTIONS} = "j$opts{j}";
+}

 # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # #
 my( $script_version ) = '2.127';

setting this option reduces my release -t on my laptop today from 65 minutes to 52 minutes. Not a huge win, but still.

2: run tests in parallel in different cloned folders
I envision more gain if I could run in a different folder per perl version 8 (or whatever) folders with a clone in parallel. This would however mean a lot more work.

If I pass a base folder like ~/test/tmp where the current folder is rsync'd too before running the tests and than clean-up when done, I can run 8 perl versions in parallel and use more of the available resources.

Sometimes I forget to update the version of the release, then various things get out of sync. I can check the current release with the MetaCPAN API and compare that to the local dist version.