cvs2cl.pl

RAM Defragmentation tools

#!/bin/sh
exec perl -w -x $0 ${1+"$@"} # -*- mode: perl; perl-indent-level: 2; -*-
#!perl -w


##############################################################
###                                                        ###
### cvs2cl.pl: produce ChangeLog(s) from `cvs log` output. ###
###                                                        ###
##############################################################

## $Revision: 1.2 $
## $Date: 2004/07/17 23:37:51 $
## $Author: knilch $
##

use strict;

use File::Basename qw( fileparse );
use Getopt::Long   qw( GetOptions );
use Text::Wrap     qw( );
use User::pwent    qw( getpwnam );

# The Plan:
#
# Read in the logs for multiple files, spit out a nice ChangeLog that
# mirrors the information entered during `cvs commit'.
#
# The problem presents some challenges. In an ideal world, we could
# detect files with the same author, log message, and checkin time --
# each <filelist, author, time, logmessage> would be a changelog entry.
# We'd sort them; and spit them out.  Unfortunately, CVS is *not atomic*
# so checkins can span a range of times.  Also, the directory structure
# could be hierarchical.
#
# Another question is whether we really want to have the ChangeLog
# exactly reflect commits. An author could issue two related commits,
# with different log entries, reflecting a single logical change to the
# source. GNU style ChangeLogs group these under a single author/date.
# We try to do the same.
#
# So, we parse the output of `cvs log', storing log messages in a
# multilevel hash that stores the mapping:
#   directory => author => time => message => filelist
# As we go, we notice "nearby" commit times and store them together
# (i.e., under the same timestamp), so they appear in the same log
# entry.
#
# When we've read all the logs, we twist this mapping into
# a time => author => message => filelist mapping for each directory.
#
# If we're not using the `--distributed' flag, the directory is always
# considered to be `./', even as descend into subdirectories.

# Call Tree

# name                         number of lines (10.xii.03)
# parse_options                         192
# derive_changelog                       13
# +-maybe_grab_accumulation_date         38
# +-read_changelog                      277
#   +-maybe_read_user_map_file           94
#     +-run_ext                           9
#   +-read_file_path                     29
#   +-read_symbolic_name                 43
#   +-read_revision                      49
#   +-read_date_author_and_state         25
#     +-parse_date_author_and_state      20
#   +-read_branches                      36
# +-output_changelog                    424
#   +-pretty_file_list                  290
#     +-common_path_prefix               35
#   +-preprocess_msg_text                30
#     +-min                               1
#   +-mywrap                             16
#   +-last_line_len                       5
#   +-wrap_log_entry                    177
#
# Utilities
#
# xml_escape                              6
# slurp_file                             11
# debug                                   5
# version                                 2
# usage                                 142

# -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*-
#
# Note about a bug-slash-opportunity:
# -----------------------------------
#
# There's a bug in Text::Wrap, which affects cvs2cl.  This script
# reveals it:
#
#   #!/usr/bin/perl -w
#
#   use Text::Wrap;
#
#   my $test_text =
#   "This script demonstrates a bug in Text::Wrap.  The very long line
#   following this paragraph will be relocated relative to the surrounding
#   text:
#
#   ====================================================================
#
#   See?  When the bug happens, we'll get the line of equal signs below
#   this paragraph, even though it should be above.";
#
#
#   # Print out the test text with no wrapping:
#   print "$test_text";
#   print "\n";
#   print "\n";
#
#   # Now print it out wrapped, and see the bug:
#   print wrap ("\t", "        ", "$test_text");
#   print "\n";
#   print "\n";
#
# If the line of equal signs were one shorter, then the bug doesn't
# happen.  Interesting.
#
# Anyway, rather than fix this in Text::Wrap, we might as well write a
# new wrap() which has the following much-needed features:
#
# * initial indentation, like current Text::Wrap()
# * subsequent line indentation, like current Text::Wrap()
# * user chooses among: force-break long words, leave them alone, or die()?
# * preserve existing indentation: chopped chunks from an indented line
#   are indented by same (like this line, not counting the asterisk!)
# * optional list of things to preserve on line starts, default ">"
#
# Note that the last two are essentially the same concept, so unify in
# implementation and give a good interface to controlling them.
#
# And how about:
#
# Optionally, when encounter a line pre-indented by same as previous
# line, then strip the newline and refill, but indent by the same.
# Yeah...

# Globals --------------------------------------------------------------------

# In case we have to print it out:
my $VERSION = '$Revision: 1.2 $';
$VERSION =~ s/\S+\s+(\S+)\s+\S+/$1/;

## Vars set by options:

# Print debugging messages?
my $Debug = 0;

# Just show version and exit?
my $Print_Version = 0;

# Just print usage message and exit?
my $Print_Usage = 0;

# What file should we generate (defaults to "ChangeLog")?
my $Log_File_Name = "ChangeLog";

# Grab most recent entry date from existing ChangeLog file, just add
# to that ChangeLog.
my $Cumulative = 0;

# `cvs log -d`, this will repeat the last entry in the old log.  This is OK,
# as it guarantees at least one entry in the update changelog, which means
# that there will always be a date to extract for the next update.  The repeat
# entry can be removed in postprocessing, if necessary.

# MJP 2003-08-02
# I don't think this actually does anything useful
my $Update = 0;

# Expand usernames to email addresses based on a map file?
my $User_Map_File = '';
my $User_Passwd_File;
my $Mail_Domain;

# Output log in chronological order? [default is reverse chronological order]
my $Chronological_Order = 0;

# Grab user details via gecos
my $Gecos = 0;

# User domain for gecos email addresses
my $Domain;

# Output to a file or to stdout?
my $Output_To_Stdout = 0;

# Eliminate empty log messages?
my $Prune_Empty_Msgs = 0;

# Tags of which not to output
my %ignore_tags;

# Show only revisions with Tags
my %show_tags;

# Don't call Text::Wrap on the body of the message
my $No_Wrap = 0;

# Indentation of log messages
my $Indent = "\t";

# Don't do any pretty print processing
my $Summary = 0;

# Separates header from log message.  Code assumes it is either " " or
# "\n\n", so if there's ever an option to set it to something else,
# make sure to go through all conditionals that use this var.
my $After_Header = " ";

# XML Encoding
my $XML_Encoding = '';

# Format more for programs than for humans.
my $XML_Output = 0;
my $No_XML_Namespace = 0;
my $No_XML_ISO_Date = 0;

# Do some special tweaks for log data that was written in FSF
# ChangeLog style.
my $FSF_Style = 0;

# Show times in UTC instead of local time
my $UTC_Times = 0;

# Show times in output?
my $Show_Times = 1;

# Show day of week in output?
my $Show_Day_Of_Week = 0;

# Show revision numbers in output?
my $Show_Revisions = 0;

# Show dead files in output?
my $Show_Dead = 0;

# Hide dead trunk files which were created as a result of additions on a
# branch?
my $Hide_Branch_Additions = 1;

# Show tags (symbolic names) in output?
my $Show_Tags = 0;

# Show tags separately in output?
my $Show_Tag_Dates = 0;

# Show branches by symbolic name in output?
my $Show_Branches = 0;

# Show only revisions on these branches or their ancestors.
my @Follow_Branches;
# Show only revisions on these branches or their ancestors; ignore descendent
# branches.
my @Follow_Only;

# Don't bother with files matching this regexp.
my @Ignore_Files;

# How exactly we match entries.  We definitely want "o",
# and user might add "i" by using --case-insensitive option.
my $Case_Insensitive = 0;

# Maybe only show log messages matching a certain regular expression.
my $Regexp_Gate = '';

# Pass this global option string along to cvs, to the left of `log':
my $Global_Opts = '';

# Pass this option string along to the cvs log subcommand:
my $Command_Opts = '';

# Read log output from stdin instead of invoking cvs log?
my $Input_From_Stdin = 0;

# Don't show filenames in output.
my $Hide_Filenames = 0;

# Don't shorten directory names from filenames.
my $Common_Dir = 1;

# Max checkin duration. CVS checkin is not atomic, so we may have checkin
# times that span a range of time. We assume that checkins will last no
# longer than $Max_Checkin_Duration seconds, and that similarly, no
# checkins will happen from the same users with the same message less
# than $Max_Checkin_Duration seconds apart.
my $Max_Checkin_Duration = 180;

# What to put at the front of [each] ChangeLog.
my $ChangeLog_Header = '';

# Whether to enable 'delta' mode, and for what start/end tags.
my $Delta_Mode = 0;
my $Delta_From = '';
my $Delta_To = '';

my $TestCode;

# Whether to parse filenames from the RCS filename, and if so what
# prefix to strip.
my $RCS_Root;

# Whether to output information on the # of lines added and removed
# by each file modification.
my $Show_Lines_Modified = 0;

## end vars set by options.

# latest observed times for the start/end tags in delta mode
my $Delta_StartTime = 0;
my $Delta_EndTime = 0;

my $No_Ancestors = 0;

my $No_Extra_Indent = 0;

my $GroupWithinDate = 0;

# ----------------------------------------------------------------------------

package CVS::Utils::ChangeLog::EntrySet;

sub new {
  my $class = shift;
  my %self;
  bless \%self, $class;
}

# -------------------------------------

sub output_changelog {
  my $output_type = $XML_Output ? 'XML' : 'Text';
  my $output_class = "CVS::Utils::ChangeLog::EntrySet::Output::${output_type}";
  my $output = $output_class->new(follow_branches => \@Follow_Branches,
                                  follow_only     => \@Follow_Only,
                                  ignore_tags     => \%ignore_tags,
                                  show_tags       => \%show_tags,
                                 );
  $output->output_changelog(@_);
}

# -------------------------------------

sub add_fileentry {
  my ($self, $file_full_path, $time, $revision, $state, $lines,
      $branch_names, $branch_roots, $branch_numbers,
      $symbolic_names, $author, $msg_txt) = @_;

      my $qunk =
        CVS::Utils::ChangeLog::FileEntry->new($file_full_path, $time, $revision,
                                              $state, $lines,
                                              $branch_names, $branch_roots,
                                              $branch_numbers,
                                              $symbolic_names);

      # We might be including revision numbers and/or tags and/or
      # branch names in the output.  Most of the code from here to
      # loop-end deals with organizing these in qunk.

      unless ( $Hide_Branch_Additions
               and
               $msg_txt =~ /file .+ was initially added on branch \S+./ ) {
        # Add this file to the list
        # (We use many spoonfuls of autovivication magic. Hashes and arrays
        # will spring into existence if they aren't there already.)

        &main::debug ("(pushing log msg for ". $qunk->dir_key . $qunk->filename . ")\n");

        # Store with the files in this commit.  Later we'll loop through
        # again, making sure that revisions with the same log message
        # and nearby commit times are grouped together as one commit.
        $self->{$qunk->dir_key}{$author}{$time}{$msg_txt} =
          CVS::Utils::ChangeLog::Message->new($msg_txt)
              unless exists $self->{$qunk->dir_key}{$author}{$time}{$msg_txt};
        $self->{$qunk->dir_key}{$author}{$time}{$msg_txt}->add_fileentry($qunk);
      }

}

# ----------------------------------------------------------------------------

package CVS::Utils::ChangeLog::EntrySet::Output::Text;

use base qw( CVS::Utils::ChangeLog::EntrySet::Output );

use File::Basename qw( fileparse );

sub new {
  my $class = shift;
  my $self = $class->SUPER::new(@_);
}

# -------------------------------------

sub wday {
  my $self = shift; my $class = ref $self;
  my ($wday) = @_;

  return $Show_Day_Of_Week ? ' ' . $class->weekday_en($wday) : '';
}

# -------------------------------------

sub header_line {
  my $self = shift;
  my ($time, $author, $lastdate) = @_;

  my $header_line = '';

  my (undef,$min,$hour,$mday,$mon,$year,$wday)
    = $UTC_Times ? gmtime($time) : localtime($time);

  my $date = $self->fdatetime($time);

  if ($Show_Times) {
    $header_line =
      sprintf "%s  %s\n\n", $date, $author;
  } else {
    if ( ! defined $lastdate or $date ne $lastdate or ! $GroupWithinDate ) {
      if ( $GroupWithinDate ) {
        $header_line = "$date\n\n";
      } else {
        $header_line = "$date  $author\n\n";
      }
    } else {
      $header_line = '';
    }
  }
}

# -------------------------------------

sub preprocess_msg_text {
  my $self = shift;
  my ($text) = @_;

  $text = $self->SUPER::preprocess_msg_text($text);