git2log

#! /usr/bin/perl

# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
#
# This script is maintained at https://github.com/openSUSE/linuxrc-devtools
#
# If you're in another project, this is just a copy.
# You may update it to the latest version from time to time...
#
# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

use strict;

use Getopt::Long;

use Data::Dumper;
$Data::Dumper::Sortkeys = 1;
$Data::Dumper::Terse = 1;
$Data::Dumper::Indent = 1;

sub usage;
sub changelog_outdated;
sub get_github_project;
sub get_version;
sub get_tags;
sub get_log;
sub is_formatted_tag;
sub get_branch;
sub choose_tags;
sub add_head_tag;
sub tags_to_str;
sub format_log;
sub format_all_logs;
sub fix_dates;
sub add_line_breaks;
sub add_bugzilla_to_weblate;
sub format_date_obs;
sub format_date_iso;
sub raw_date_to_s;

usage 0 if !@ARGV;

my @changelog_deps = qw ( .git/HEAD .git/refs/heads .git/refs/tags );

my $branch;
my $current_version;
my @tags;
my @all_tags;
my $config;

my $opt_log;
my $opt_version;
my $opt_branch;
my $opt_update;
my $opt_file;
my $opt_start;
my $opt_max;
my $opt_width = 66;
my $opt_width_fuzz = 8;
my $opt_sep_width = 68;
my $opt_format = 'internal';		# obs, internal
my $opt_merge_msg_before = 1;		# log auto generated pr merge message before the commit messages (vs. after)
my $opt_join_author = 1;		# join consecutive commit messages as long as they are by the same author
my $opt_keep_date = 1;			# don't join consecutive commit messages if they have different time stamps
my $opt_default_email = 'opensuse-packaging@opensuse.org';	# default email to use in changelog
my $opt_weblate = 'bsc#1149754';	# bugzilla ref to use for Weblate commits

$opt_default_email = $ENV{GIT2LOG_DEFAULT_EMAIL} if $ENV{GIT2LOG_DEFAULT_EMAIL};

GetOptions(
  'help'          => sub { usage 0 },
  'version'       => \$opt_version,
  'branch'        => \$opt_branch,
  'update'        => \$opt_update,
  'start=s'       => \$opt_start,
  'format=s'      => \$opt_format,
  'max=i'         => \$opt_max,
  'width=i'       => \$opt_width,
  'fuzz=i'        => \$opt_width_fuzz,
  'merge-msg=s'   => sub { $opt_merge_msg_before = ($_[1] eq 'after' ? 0 : 1) },
  'join-author!'  => \$opt_join_author,
  'keep-date!'    => \$opt_keep_date,
  'log|changelog' => \$opt_log,
  'default-email=s' => \$opt_default_email,
  'weblate=s'     => \$opt_weblate,
) || usage 1;

# ensure we are used correctly
usage 1 if @ARGV > 1 || !($opt_log || $opt_version || $opt_branch);
$opt_file = @ARGV ? shift : '-';

die "no git repo\n" unless -d ".git";

# if update option has been give write changelog only if git refs are newer
exit 0 if $opt_update && $opt_file ne '-' && -f($opt_file) && !changelog_outdated($opt_file);

$opt_max = 2 if $opt_version || $opt_branch;

# gather some data
get_github_project;
get_branch;
get_log;
fix_dates;
get_tags;
choose_tags;
add_head_tag;
get_version;

# just print current branch
if($opt_branch) {
  open my $f, ">$opt_file";
  print $f $config->{branch} ? $config->{branch} : "master", "\n";
  close $f;

  exit 0;
}

# just print current version
if($opt_version) {
  my $old_version;

  if($opt_file ne '-' && open(my $f, $opt_file)) {
    chomp($old_version = <$f>);
    close $f;
  }

  if($config->{version} ne $old_version) {
    open my $f, ">$opt_file";
    print $f "$config->{version}\n";
    close $f;
  }

  exit 0;
}

# set start tag
if($opt_start) {
  my $x = is_formatted_tag $opt_start;
  die "$opt_start: not a valid start tag\n" if !$x;
  $x->{branch} = $config->{branch} if !$x->{branch};
  $config->{start} = $x;
}

format_all_logs;

open my $f, ">$opt_file";

print $f $_->{formatted} for @{$config->{log}};

close $f;

exit 0;


# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# usage(exit_code)
#
# Print help message and exit.
# - exit_code: exit code
#
# Function does not return.
#
sub usage
{
  my $err = shift;

  print <<"  usage";
Usage: git2log [OPTIONS] [FILE]
Create changelog and project version from git repo.
  --changelog         Write changelog to FILE.
  --version           Write version number to FILE.
  --branch            Write current branch to FILE.
  --start START_TAG   Start with tag START_TAG.
  --max N             Write at most N log entries.
  --update            Write changelog or version only if FILE is outdated.
  --format FORMAT     Write log using FORMAT. Supported FORMATs are 'internal' (default) and 'obs'.
  --width WIDTH       Reformat log entries to be max WIDTH chars wide.
  --fuzz FUZZ         Allow log lines to be up to FUZZ chars longer as WIDTH to avoid
                      line breaks leaving tiny bits on the last line.
  --merge-msg WHERE   Log message about merges before or after the actual merge commit messages.
                      Valid values for WHERE are 'after' and 'before' (default).
  --join-author       Join consecutive commits as long as they are by the same author. (default)
  --no-join-author    Keep consecutive commits by the same author separate.
  --keep-date         Join consecutive commits only if they have the same date. (default)
  --no-keep-date      Join consecutive commits even if dates differ.
  --default-email     Use this email in changelog entries if no other suitable email could be
                      determined (default: opensuse-packaging\@opensuse.org).
  --weblate STRING    Add this STRING to every auto-generated Weblate commit.
  --help              Print this help text.
  usage

  exit $err;
}


# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# res = changelog_outdated(file)
#
# Return status of changelog file.
# - file: changelog file name
# - res: status
#     1: file is newer than the last git repo change and should be updated
#     0: file is still recent enough
#
# Relies on global var @changelog_deps.
#
sub changelog_outdated
{
  my $file = $_[0];

  my $changelog_time = (stat $file)[9];

  return 1 if !defined $changelog_time;

  for (@changelog_deps) {
    return 1 if (stat)[9] > $changelog_time;
  }

  return 0;
}


# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# get_github_project()
#
# Set $config->{github_project} to the github project name.
#
sub get_github_project
{
  if(`git config remote.origin.url` =~ m#github.com[:/]+(\S+/\S+)#) {
    $config->{github_project} = $1;
    $config->{github_project} =~ s/\.git$//;
  }
}


# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# get_version()
#
# Set $config->{branch} and $config->{version} to the current branch and
# version info.
#
# This might be taken directly from HEAD if HEAD is tagged or otherwise be
# exprapolated from the most recent tag (cf. add_head_tag()).
#
sub get_version
{
  $config->{version} = "0.0";

  my $tag = $config->{log}[0]{tags}[0];

  if($tag->{version}) {
    $config->{version} = $tag->{version};
    $config->{branch} = $tag->{branch};
  }
}


# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# get_tags()
#
# Parse $config->{raw_log}, extract tag names, and split into per-tag
# sections.
#
# Only tags recognized by is_formatted_tag() are considered.
#
# Tags inside merge commits are ignored.
#
# The parsed logs is stored in $config->{log}, an array of log sections.
# Each section is a hash with these keys:
#   - 'tags': array of tags for this section
#   - 'commit': git commit id associated with these tags
#   - 'lines': git log lines
#
sub get_tags
{
  my $pre_parsed;
  my $current;

  for (@{$config->{raw_log}}) {
    if(/^commit (\S+)( \((.*)\))?/) {
      $current = { commit => $1, tag_list => $3 };
      push @$pre_parsed, $current;
    }

    next if !defined $current;

    if(/^Merge: (\S+) (\S+)/) {
      $current->{merge_end} = $1;
      $current->{merge_start} = $2;
    }

    push @{$current->{lines}}, $_;
  }

  # the end of the merge commit if in a merge
  my $merge;
  my $log_entry;

  for my $p (@$pre_parsed) {
    my $xtag;

    # we have reached the end of the merge commit
    if((defined($merge) && $p->{commit} =~ /^$merge/) || (!defined($merge) && $p->{merge_end})) {
      undef $merge;
    }

    # ignore tag info inside a merge commit
    if(!$merge) {
      for my $t (split /, /, $p->{tag_list}) {
        if($t =~ /tag: (\S+)/) {
          my $tag = $1;
          my $x = is_formatted_tag $tag;
          push @$xtag, $x if $x;
        }
      }

      if($p->{merge_end}) {
        # remember end of current merge commit
        $merge = $p->{merge_end};
      }
    }

    if(!$log_entry) {
      $log_entry = { commit => $p->{commit} };
      $log_entry->{tags} = $xtag if $xtag;
      push @{$config->{log}}, $log_entry;
    }
    else {
      if($xtag) {
        $log_entry = { commit => $p->{commit}, tags => $xtag };
        push @{$config->{log}}, $log_entry;
      }
    }

    push @{$log_entry->{lines}}, @{$p->{lines}};

    last if $opt_max && @{$config->{log}} >= $opt_max;
  }
}


# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# get_log()
#
# Read git log and store lines as array in $config->{raw_log} (trailing
# newlines removed).
#
sub get_log
{
  chomp(@{$config->{raw_log}} = `git log --pretty=medium --date=raw --topo-order --decorate`);
}


# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# hash_ref = is_formatted_tag(tag_name)
#
# Parse tag and return hash ref with branch and version number parts or
# undef if it doesn't match.
# - tag_name: tag as string
# - hash_ref: hash ref with internal tag representation (with keys 'branch' and 'version').
#
# This expects tags of the form "VERSION" or "BRANCH-VERSION" where VERSION
# consists of decimal numbers separated by dots '.' and BRANCH can be any
# string.
# (Note: it doesn't really have to be the name of an existing branch.)
#
# Tags not conforming to this convention are ignored.
#
sub is_formatted_tag
{
  if($_[0] =~ /^((.+)-)?((\d+\.)*\d+)$/) {
    return { branch => $2, version => $3 }
  }

  return undef;
}


# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# get_branch()
#
# Get currently active git branch and store in $config->{branch}.
#
# 'master' branch is represented by empty 'branch' key.
#
sub get_branch
{
  chomp(my $branch = `git rev-parse --abbrev-ref HEAD`);

  $branch = "" if $branch eq 'master';

  $config->{branch} = $branch;
}


# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# res = tag_sort(a, b)
#
# Compare 2 tags.
# - a, b: refs to tag hash
# - res: -1, 0, 1
#
# This is used when we have to decide between alternative tags.
# (Prefer 'lesser' variant.)
#
sub tag_sort
{
  my ($x, $y);

  $x = length $a->{version};
  $y = length $b->{version};

  # longer version number first
  return $y <=> $x if $y <=> $x;

  return $a->{branch} cmp $b->{branch};
}


# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# str = tag_to_str(tag_ref)
#
# Convert tag into string.
# - tag_ref: ref to hash with 'branch' and 'version' keys
# - str: string (e.g. "foo-1.44")
#
# 'master' branch is represented by missing/empty 'branch' key.
#
sub tag_to_str
{
  my $tag = $_[0];
  my $str;

  $str = "$tag->{branch}-" if $tag->{branch} ne "";
  $str .= $tag->{version};

  return $str;
}


# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# str = tags_to_str(tag_array_ref)
#
# Convert array of tags into string.
# - tag_array_ref: ref to array of tags
# - str: string (e.g. "(tag1, tag2)"
#
# This function is used only internally for debugging.
#
sub tags_to_str
{
  my $tags = $_[0];
  my $str;

  for my $t (@$tags) {
    $str .= ", " if $str;
    $str .= tag_to_str $t;
  }

  return "($str)";
}


# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# choose_tags()
#
# Scan commit messages and extract tag & branch information.
#
# This stores the tag/branch info in $config->{log}[]{tags}.
#
sub choose_tags
{
  my $branch = $config->{branch};

  for my $x (@{$config->{log}}) {
    # printf "# %s\n", tags_to_str($x->{tags});

    # no tag info? -> ignore
    next if !$x->{tags};

    # single tag? -> remember branch info
    if(@{$x->{tags}} == 1) {
      $branch = $x->{tags}[0]{branch};
      next;
    }

    # several tags? -> choose one

    # any with current branch name?
    my @t = grep { $_->{branch} eq $branch } @{$x->{tags}};

    # no? -> choose among all
    @t = @{$x->{tags}} if @t == 0;

    # prefer longest version number, then alphanumerically smallest branch name
    @t = sort tag_sort @t;

    $branch = $t[0]{branch};

    # Here's some magic:
    #
    # If a commit is tagged  "FOO-X.Y" *and* "FOO-X.Y.0" assume the user
    # wants to start a new sub-numbering scheme.
    #
    # Use "X.Y" as version but remember to go to "X.Y.1" instead of
    # "X.{Y+1}" when we need to increment the version.
    #
    my $version = $t[0]{version};
    if($version =~ s/\.0$//) {
      if(grep { $_->{branch} eq $branch && $_->{version} eq $version } @t) {
        $t[0]{new_start} = 1;
        $t[0]{version} = $version;
      }
    }

    $x->{tags} = [ $t[0] ];

    # printf "X %s\n", tags_to_str($x->{tags});
  }
}


# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# add_head_tag()
#
# Suggest tag for HEAD if there isn't one.
#
# Basically, use branch + version from most recent tag and increment version.
#
# If the 'new_start' attribute to tag is set, start a new sub-numbering scheme.
#
sub add_head_tag
{
  return if @{$config->{log}} < 2;

  # HEAD tagged already?
  return if $config->{log}[0]{tags};

  # the first tagged commit if HEAD isn't tagged
  my $tag = { %{$config->{log}[1]{tags}[0]} };

  # append '.0' to version
  $tag->{version} .= '.0' if $tag->{new_start};

  # increment version
  $tag->{version} =~ s/(\d+)$/$1 + 1/e;

  $config->{log}[0]{tags}[0] = $tag;

  # remember that the tag was generated
  $config->{log}[0]{was_untagged} = 1;
}


# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# fix_dates()
#
# Adjust time stamps in entire git log.
#
# The time stamps of the git commits are not necessarily ordered by date.
# But the generated changelog is required to have a strictly monotonic time.
#
# We do this by going through the log in reverse and rewriting any dates we
# find whenever the date decreases.
#
# A minimum time difference of 1 second beween entries is maintained.
#
# Not very subtle but it works.
#
sub fix_dates
{
  my $last_date;

  for (reverse @{$config->{raw_log}}) {
    # e.g. "Date: 1443184889 +0200"
    if(/^(Date:\s+)(\S+)(\s+\S+)/) {
      if(defined $last_date && $2 < $last_date) {
        $_ = "$1$last_date$3\n";
      }
      else {
        $last_date = $2;
      }

      # ensure a minimal time gap of 1 second
      $last_date += 1;
    }
  }
}


# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# format_all_logs()
#
# Format the entire git log.
#
# This is done for every code version individually (the log has already been
# split accordingly).
#
# If $config->{start} is set, use this as starting point. Else format the
# entire git log.
#
sub format_all_logs
{
  # check if start tag actually exists - if not, print nothing
  if($config->{start}) {
    my $tag_found;
    for (@{$config->{log}}) {
      $tag_found = 1, last if grep { tag_to_str($config->{start}) eq tag_to_str($_) } @{$_->{tags}};
    }
    return if !$tag_found;
  }

  for (@{$config->{log}}) {
    if($config->{start}) {
      # stop if we meet the start tag
      last if grep { tag_to_str($config->{start}) eq tag_to_str($_) } @{$_->{tags}};
    }
    format_log $_;
  }
}


# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# format_log(log)
#
# Format log messages.
# - log: is an array ref with individual commits
#
# All commits belong to a specific code version (stored in $log->{tag}).
# $log->{formatted} holds the result.
#
# The process is done in several individual steps, documented below in the code.
#
sub format_log
{
  my $log = $_[0];

  my $merge;
  my $commit;
  my $commits;

  for (@{$log->{lines}}) {
    if(/^commit (\S+)/) {
      $commit = { ref => $1 };
      push @{$commits}, $commit;

      if(
        $merge &&
        $merge->{merge_end} eq substr($commit->{ref}, 0, length($merge->{merge_end}))
      ) {
        undef $merge;
      }

      if($merge) {
        $commit->{merge_ref} = $merge->{ref};
        $commit->{date} = $merge->{date};
        $commit->{author} = $merge->{author};
        # add to all commits so it's not lost when we re-arrange
        $commit->{merge_msg} = $merge->{msg};
      }

      next;
    }

    if(/^Merge: (\S+)/ && !$merge) {
      if($commit) {
        $merge = { merge_end => $1, ref => $commit->{ref} } unless $merge;
      }
      next;
    }

    if(/^Date:\s+(\S.*)/) {
      $commit->{date} ||= $1 if $commit;
      $merge->{date} ||= $1 if $merge;
      next;
    }

    if(/^Author:\s+(\S.*)/) {
      $commit->{author} ||= $1 if $commit;
      $merge->{author} ||= $1 if $merge;
      next;
    }

    if($merge) {
      if(/^    Merge pull request (#\d+) from (\S+)/) {
        if($config->{github_project}) {
          push @{$merge->{msg}}, "merge gh#$config->{github_project}$1";
        }
        else {
          push @{$merge->{msg}}, "merge pr $2";
        }
      }
      elsif(/^    Merge branch '([^']+)'( into)?/) {
        push @{$merge->{msg}}, "merge branch $1" if $2 eq "";
      }
      elsif(/^    Merge remote-tracking branch /) {
        # ignore
      }
      elsif(s/^    //) {
        push @{$commit->{lines}}, $_ unless /^# /;
      }
    }
    elsif($commit) {
      if(s/^    //) {
        push @{$commit->{lines}}, $_ unless /^# /;
      }
    }
  }

  # Note: the individual steps below work on the array @$commits and modify
  # its content.

  # step 1
  # - if there are paragraphs starting with '@log@' or '@+log@'
  #     - delete first paragraph (short summary)
  # - else
  #     - keep only first paragraph
  # - if there is a paragraph starting with '@-log', delete entire log
  # - tag commits that have a '@log@' tag so we can delete untagged commits
  #   belonging to the same merge commit later

  my $tagged_merges = {};

  for my $commit (@$commits) {
    # skip leading empty lines
    for (@{$commit->{lines}}) {
      last if !/^\s*$/;
      shift @{$commit->{lines}};
    }
    my $para_cnt = 0;
    my $delete_all = 0;
    my $delete_first = 0;
    for (@{$commit->{lines}}) {
      $para_cnt++ if $_ eq "";
      $para_cnt = 0, $delete_first = 1 if /^\@\+log\@/;
      $delete_all = 1 if /^\@\-log\@/;
      if(/^\@log\@/) {
        $para_cnt = 0;
        $commit->{clear} = 1;
        $tagged_merges->{$commit->{merge_ref}} = 1 if $commit->{merge_ref} || $log->{was_untagged};
      }
      $_ = undef if $para_cnt;
    }
    shift @{$commit->{lines}} if $delete_first;
    $commit->{lines} = [] if $delete_all;
  }

  # step 2
  # - clean up tagged commits or commits belonging to tagged merges

  for my $commit (@$commits) {
    next unless $commit->{clear} || $tagged_merges->{$commit->{merge_ref}};
    for (@{$commit->{lines}}) {
      last if /^\@\+?log\@/;
      $_ = undef;
    }
  }

  # step 3
  # - join lines

  for my $commit (@$commits) {
    my $lines;
    my $line;

    for (@{$commit->{lines}}) {
      next if $_ eq "";
      if(
        s/^\s*[+\-][\-\s]*// ||
        s/^\@\+?log\@// ||
        $line eq ""
      ) {
        s/^\s*//;
        push @$lines, $line if $line ne "";
        $line = $_;
      }
      else {
        s/^\s*//;
        $line .= " " if $line ne "";
        $line .= $_;
      }
    }
    push @$lines, $line if $line ne "";

    $commit->{formatted} = $lines if $lines;
  }

  # step 4
  # - fix small glitches

  for my $commit (@$commits) {
    next unless $commit->{formatted};
    for (@{$commit->{formatted}}) {
      s/(fate|bnc|bsc|boo|jsc)\s*(#[a-z\d\-]+)/\L$1\E$2/ig;
    }
  }

  # step 5
  # - add merge info at the top or bottom (depending on $opt_merge_msg_before)

  my $merge_logged;

  for my $commit ($opt_merge_msg_before ? reverse(@$commits) : @$commits) {
    next unless $commit->{formatted};

    if($commit->{merge_ref} && !$merge_logged->{$commit->{merge_ref}}) {
      $merge_logged->{$commit->{merge_ref}} = 1;
      if($commit->{merge_msg}) {
        if($opt_merge_msg_before) {
          unshift @{$commit->{formatted}}, @{$commit->{merge_msg}};
        }
        else {
          push @{$commit->{formatted}}, @{$commit->{merge_msg}};
        }
      }
    }
  }

  # step 6
  # - join commit messages with same author (optionally even with different dates)

  my $commit0;

  for my $commit (@$commits) {
    next if !$commit->{formatted};
    $commit0 = $commit, next if !$commit0;

    if(
      # $commit->{merge_ref} eq $commit0->{merge_ref} &&
      (
        $opt_join_author && ($commit->{author} eq $commit0->{author})
        && (!$opt_keep_date || $commit->{date} eq $commit0->{date})
      )
      || $opt_format eq 'internal'
    ) {
      unshift @{$commit0->{formatted}}, @{$commit->{formatted}};
      delete $commit->{formatted};
    }
    else {
      $commit0 = $commit;
    }
  }

  # step 7
  # - add version tag at the end of the first log entry

  for my $commit (@$commits) {
    next unless $commit->{formatted};

    if($opt_format eq 'obs') {
      push @{$commit->{formatted}}, $log->{tags}[0]{version} if defined $log->{tags}[0]{version};
    }
    else {
      # push @{$commit->{formatted}}, tag_to_str($log->{tags}[0]);
    }

    last;
  }

  # step 8
  # - remove identical lines

  for my $commit (@$commits) {
    next unless $commit->{formatted};
    my %k;
    $commit->{formatted} = [ grep { !$k{$_}++ } @{$commit->{formatted}} ]
  }

  # step 9
  # - remove shortened lines (that match the beginning of other lines)

  for my $commit (@$commits) {
    next unless $commit->{formatted};

    # return 1 if some other commit line starts with function arg
    my $is_substr = sub {
      my $str = $_[0];
      $str =~ s/\s*…$//;	# github likes to shorten lines with ' …'
      my $str_len = length $str;
      for (@{$commit->{formatted}}) {
        return 1 if $str_len < length($_) && $str eq substr($_, 0, $str_len);
      }

      return 0;
    };

    $commit->{formatted} = [ grep { ! $is_substr->($_) } @{$commit->{formatted}} ]
  }

  # step 10
  # - add bugzilla reference to Weblate commits

  for my $commit (@$commits) {
    next unless $commit->{formatted};
    for (@{$commit->{formatted}}) {
      $_ = add_bugzilla_to_weblate $_;
    }
  }

  # step 11
  # - add line breaks

  for my $commit (@$commits) {
    next unless $commit->{formatted};
    for (@{$commit->{formatted}}) {
      $_ = add_line_breaks $_;
    }
  }

  # step 12
  # - generate final log message
  #
  # note: non-(open)suse email addresses are replaced by $opt_default_email

  my $formated_log;

  for my $commit (@$commits) {
    next unless $commit->{formatted} && @{$commit->{formatted}};

    if($opt_format eq 'obs') {
      $formated_log .= "-" x $opt_sep_width . "\n";
      $formated_log .= format_date_obs($commit->{date});
    }
    else {
      $formated_log .= format_date_iso($commit->{date});
    }
    if($opt_format eq 'obs') {
      my $auth = $commit->{author};
      $auth =~ s/^.*<//;
      $auth =~ s/>.*$//;
      # replace non-suse e-mail addresses with a generic one
      if($auth !~ /\@(suse\.(com|cz|de)|opensuse\.org)$/) {
        $auth = $opt_default_email;
      }
      $formated_log .= " - $auth\n\n";
    }
    else {
      $formated_log .= ":\t" . tag_to_str($log->{tags}[0]) . "\n";
    }

    for (@{$commit->{formatted}}) {
      s/^/\t/mg if $opt_format eq 'internal';
      $formated_log .= "$_\n";
    }

    $formated_log .= "\n";
  }

  $log->{formatted} = $formated_log;
}


# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# new_text = add_line_breaks(text)
#
# Add line breaks to text.
# - text: some text
# - new_text: same text, reformatted
#
# Lines are formatted to have a maximal length of $opt_width. If this causes
# the last line to be shorter than $opt_width_fuzz, it is appended to the
# previous line.
#
sub add_line_breaks
{
  my @words = split /\s+/, @_[0];
  my $remaining_len = length(join '', @words);

  my $str = shift(@words);
  my $len = length $str;

  my $next_len;
  my $word_len;

  for (@words) {
    $word_len = length;
    $remaining_len -= $word_len;
    $next_len = $len + $word_len + 1;
    if(
      $next_len >= $opt_width &&
      $next_len + $remaining_len + 1 >= $opt_width + $opt_width_fuzz
    ) {
      $str .= "\n  $_";
      $len = $word_len;
    }
    else {
      $str .= " $_";
      $len += $word_len + 1;
    }
  }

  return "- " . $str;
}


# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# new_text = add_bugzilla_to_weblate(text)
#
# Add bugzilla number to an auto-generated Weblate commit.
# - text: some text
# - new_text: same text, "($opt_weblate)" added
#
sub add_bugzilla_to_weblate
{
  my $text = @_[0];

  if($opt_weblate ne "") {
    if($text =~ /Translated using Weblate/ && $text !~ /\($opt_weblate\)/) {
      $text .= " ($opt_weblate)";
    }
  }

  return $text;
}


# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# seconds = raw_date_to_s(git_date)
#
# Convert git raw date to seconds.
# - git_date: raw git format (e.g. "1443184889 +0200")
# - seconds: the seconds part (e.g. "1443184889")
#
sub raw_date_to_s
{
  return (split / /, $_[0])[0];
}


# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# date = format_date_obs(git_date)
#
# Convert git raw date to obs format.
# - git_date: raw git format (e.g. "1443184889 +0200")
# - date: obs format ("Fri Sep 25 12:41:29 UTC 2015")
#
sub format_date_obs
{
  my @d = gmtime(raw_date_to_s($_[0]));

  return
    qw(Sun Mon Tue Wed Thu Fri Sat)[$d[6]] . " " .
    qw(Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec)[$d[4]] . " " .
    $d[3] . " " .
    sprintf("%02d:%02d:%02d", $d[2], $d[1], $d[0]) . " UTC " .
    (1900 + $d[5]);
}


# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# date = format_date_iso(git_date)
#
# Convert git raw date to iso format.
# - git_date: raw git format (e.g. "1443184889 +0200")
# - date: obs format ("2015-09-25")
#
sub format_date_iso
{
  my @d = gmtime(raw_date_to_s($_[0]));

  return sprintf("%04d-%02d-%02d", 1900 + $d[5], $d[4] + 1, $d[3]);
}