annotate lib-src/rcs2log @ 636:dce8cdbac0ea

Ignore log messages that start with `#'.

#!/bin/sh

# RCS to ChangeLog generator

#	$Id: rcs2log,v 1.5 1992/04/01 08:57:55 eggert Exp eggert $

# Generate a change log prefix from RCS/* and the existing ChangeLog (if any).
# Output the new prefix to standard output.
# You can edit this prefix by hand, and then prepend it to ChangeLog.

# Ignore log entries that start with `#'.
# Clump together log entries that start with `{topic} ',
# where `topic' contains neither white space nor `}'.

# Parse options.

# defaults
indent=8 # indent of log line
length=79 # suggested max width of log line
tabwidth=8 # width of horizontal tab

while :
do
	case $1 in
	-i)	indent=${2?};;
	-l)	length=${2?};;
	-t)	tabwidth=${2?};;
	-*)	echo >&2 "$0: usage: $0 [-i indent] [-l length] [-t tabwidth] [file ...]"
		exit 1;;
	*)	break
	esac
	shift; shift
done


# Log into $rlogout the revisions checked in since the first ChangeLog entry.

date=1970
if test -s ChangeLog
then
	# Add 1 to seconds to avoid duplicating most recent log.
	# It's a good thing `rlog' doesn't mind a time ending in `:60'.
	e='
		/^... ... [ 0-9][0-9] [ 0-9][0-9]:[0-9][0-9]:[0-9][0-9] [0-9]+ /{
			printf "%s%.2d %s\n", substr($0,1,17), substr($0,18,2)+1, $5
			exit
		}
	'
	d=`awk "$e" <ChangeLog` || exit
	case $d in
	?*) date=$d
	esac
fi
datearg="-d>$date"

rlogout=/tmp/chg$$
trap exit 1 2 13 15
trap 'rm -f $rlogout; exit 1' 0

case $# in
0) set RCS/*
esac

rlog "$datearg" "$@" >$rlogout || exit


# Get the full name of each author the logs mention, and set initialize_fullname
# to awk code that initializes the `fullname' awk associative array.
# Warning: foreign authors (i.e. not known in the passwd file) are mishandled;
# you have to fix the resulting output by hand.

authors=`
	sed -n 's|^date: *[0-9]*/[0-9][0-9]/[0-9][0-9] [0-9][0-9]:[0-9][0-9]:[0-9][0-9]; *author: *\([^; ]*\).*|\1|p' <$rlogout |
	sort -u
`

initialize_fullname=
for author in $authors
do
	fullname=`
		(grep "^$author:" /etc/passwd || ypmatch "$author" passwd) |
		sed -n 's/^[^:]*:[^:]*:[^:]*:[^:]*:\([^,:]*\).*$/\1/;p;q'
	`
	case $fullname in
	*\&*)
		User=`
			expr " $author" : ' \(.\)' |
			tr abcdefghijklmnopqrstuvwxyz ABCDEFGHIJKLMNOPQRSTUVWXYZ
		``
			expr " $author" : ' .\(.*\)'
		`
		fullname=`echo "$fullname" | sed "s:&:$User:"`
	esac
	case $fullname in
	?*)
		initialize_fullname="$initialize_fullname
			fullname[\"$author\"] = \"$fullname\""
	esac
done


# Function to print a single log line.
# We don't use awk functions, to stay compatible with old awk versions.
# `Log' is the log message (with \n replaced by \r).
# `files' contains the affected files.
printlogline='{

	# Following the GNU coding standards, rewrite
	#	* file: (function): comment
	# to
	#	* file (function): comment
	if (Log ~ /^\([^)]*\): /) {
		i = index(Log, ")")
		files = files " " substr(Log, 1, i)
		Log = substr(Log, i+3)
	}

	# If "label: comment" is too long, break the line after the ":".
	sep = " "
	if ('"$length"' <= '"$indent"' + 1 + length(files) + index(Log, "\r")) sep = "\n" indent_string

	# Print the label.
	printf "%s*%s:", indent_string, files

	# Print each line of the log, transliterating \r to \n.
	while ((i = index(Log, "\r")) != 0) {
		printf "%s%s\n", sep, substr(Log, 1, i-1)
		sep = indent_string
		Log = substr(Log, i+1)
	}
}'

hostname=`(
	hostname || cat /etc/whoami || uuname -l || uname -n
) 2>/dev/null` || {
	echo >&2 "$0: cannot deduce hostname"
	exit 1
}


# Process the rlog output, generating ChangeLog style entries.

# First, reformat the rlog output so that each line contains one log entry.
# Transliterate \n to \r so that multiline entries fit on a single line.
# Discard irrelevant rlog output.
awk <$rlogout '
	/^Working file:/ { filename = $3 }
	/^date: /, /^(-----------*|===========*)$/ {
		if ($0 ~ /^branches: /) { next }
		if ($0 ~ /^date: [0-9][ /0-9:]*;/) {
			time = substr($3, 1, length($3)-1)
			author = substr($5, 1, length($5)-1)
			printf "%s %s %s %s \r", filename, $2, time, author
			next
		}
		if ($0 ~ /^(-----------*|===========*)$/) { print ""; next }
		printf "%s\r", $0
	}
' |

# Now each line is of the form
# FILENAME YYYY/MM/DD HH:MM:SS AUTHOR \rLOG
#	where \r stands for a carriage return,
#	and each line of the log is terminated by \r instead of \n.
# Sort the log entries, first by date (in reverse order),
# then by author, then by log entry, and finally by file name (just in case).
sort +1 -3r +3 +0 |

# Finally, reformat the sorted log entries.
awk '
	BEGIN {

		# Initialize the fullname associative array.
		'"$initialize_fullname"'

		# Initialize indent string.
		indent_string = ""
		i = '"$indent"'
		if (0 < '"$tabwidth"')
			for (;  '"$tabwidth"' <= i;  i -= '"$tabwidth"')
				indent_string = indent_string "\t"
		while (1 <= i--)
			indent_string = indent_string " "

		# Set up date conversion tables.
		# RCS uses a nice, clean, sortable format,
		# but ChangeLog wants the traditional, ugly ctime format.

		# January 1, 0 AD (Gregorian) was Saturday = 6
		EPOCH_WEEKDAY = 6
		# Of course, there was no 0 AD, but the algorithm works anyway.

		w[0]="Sun"; w[1]="Mon"; w[2]="Tue"; w[3]="Wed"
		w[4]="Thu"; w[5]="Fri"; w[6]="Sat"

		m[0]="Jan"; m[1]="Feb"; m[2]="Mar"
		m[3]="Apr"; m[4]="May"; m[5]="Jun"
		m[6]="Jul"; m[7]="Aug"; m[8]="Sep"
		m[9]="Oct"; m[10]="Nov"; m[11]="Dec"

		# days in non-leap year thus far, indexed by month (0-12)
		mo[0]=0; mo[1]=31; mo[2]=59; mo[3]=90
		mo[4]=120; mo[5]=151; mo[6]=181; mo[7]=212
		mo[8]=243; mo[9]=273; mo[10]=304; mo[11]=334
		mo[12]=365
	}

	{
		newlog = substr($0, 1 + index($0, "\r"))

		# Ignore log entries prefixed by "#".
		if (newlog ~ /^#/) { next }

		if (Log != newlog || date != $2 || author != $4) {

			# The previous log and this log differ.

			# Print the old log.
			if (date != "") '"$printlogline"'

			# Logs that begin with "{clumpname} " should be grouped together,
			# and the clumpname should be removed.
			# Extract the new clumpname from the log header,
			# and use it to decide whether to output a blank line.
			newclumpname = ""
			sep = "\n"
			if (date == "") sep = ""
			if (newlog ~ /^{[^	 }]*}[	 ]/) {
				i = index(newlog, "}")
				newclumpname = substr(newlog, 1, i)
				while (substr(newlog, i+1) ~ /^[	 ]/) i++
				newlog = substr(newlog, i+1)
				if (clumpname == newclumpname) sep = ""
			}
			printf sep
			clumpname = newclumpname

			# Get ready for the next log.
			Log = newlog
			if (files != "")
				for (i in filesknown)
					filesknown[i] = 0
			files = ""
		}
		if (date != $2  ||  author != $4) {
			# The previous date+author and this date+author differ.
			# Print the new one.
			date = $2
			author = $4

			# Convert nice RCS date like "1992/01/03 00:03:44"
			# into ugly ctime date like "Fri Jan  3 00:03:44 1992".
			# Calculate day of week from Gregorian calendar.
			i = index($2, "/")
			year = substr($2, 1, i-1) + 0
			monthday = substr($2, i+1)
			i = index(monthday, "/")
			month = substr(monthday, 1, i-1) + 0
			day = substr(monthday, i+1) + 0
			leap = 0
			if (2 < month && year%4 == 0 && (year%100 != 0 || year%400 == 0)) leap = 1
			days_since_Sunday_before_epoch = EPOCH_WEEKDAY + year * 365 + int((year + 3) / 4) - int((year + 99) / 100) + int((year + 399) / 400) + mo[month-1] + leap + day - 1

			# Print "date  fullname  (email address)" if the fullname is known;
			# print "date  author" otherwise.
			# Get the fullname from the associative array.
			# The email address is just author@thishostname.
			printf "%s %s %2d %s %d  ", w[days_since_Sunday_before_epoch%7], m[month-1], day, $3, year
			if (fullname[author])
				printf "%s  (%s@%s)\n\n", fullname[author], author, "'"$hostname"'"
			else
				printf "%s\n\n", author
		}
		if (! filesknown[$1]) {
			filesknown[$1] = 1
			if (files == "") files = " " $1
			else files = files ", " $1
		}
	}
	END {
		# Print the last log.
		if (date != "") {
			'"$printlogline"'
			printf "\n"
		}
	}
' &&


# Exit successfully.

exec rm -f $rlogout