mirror of
https://github.com/mozilla/gecko-dev.git
synced 2024-11-08 12:37:37 +00:00
732 lines
18 KiB
Perl
Executable File
732 lines
18 KiB
Perl
Executable File
# -*- Mode: perl; indent-tabs-mode: nil -*-
|
|
#
|
|
# The contents of this file are subject to the Netscape Public
|
|
# License Version 1.1 (the "License"); you may not use this file
|
|
# except in compliance with the License. You may obtain a copy of
|
|
# the License at http://www.mozilla.org/NPL/
|
|
#
|
|
# Software distributed under the License is distributed on an "AS
|
|
# IS" basis, WITHOUT WARRANTY OF ANY KIND, either express or
|
|
# implied. See the License for the specific language governing
|
|
# rights and limitations under the License.
|
|
#
|
|
# The Original Code is the Tinderbox build tool.
|
|
#
|
|
# The Initial Developer of the Original Code is Netscape Communications
|
|
# Corporation. Portions created by Netscape are
|
|
# Copyright (C) 1998 Netscape Communications Corporation. All
|
|
# Rights Reserved.
|
|
#
|
|
# Contributor(s):
|
|
|
|
# Reading the log backwards saves time when we only want the tail.
|
|
use Backwards;
|
|
|
|
#
|
|
# Global variabls and functions for tinderbox
|
|
#
|
|
|
|
#
|
|
# Global variables
|
|
#
|
|
|
|
# From load_data()
|
|
$ignore_builds = {};
|
|
$scrape_builds = {};
|
|
|
|
# From get_build_name_index()
|
|
$build_name_index = {};
|
|
$build_names = [];
|
|
$name_count = 0;
|
|
|
|
# Frome get_build_time_index()
|
|
$build_time_index = {};
|
|
$build_time_times = [];
|
|
$mindate_time_count = 0; # time_count that corresponds to the mindate
|
|
$time_count = 0;
|
|
|
|
$build_table = [];
|
|
$who_list = [];
|
|
@note_array = ();
|
|
|
|
$gzip = '/usr/bin/gzip';
|
|
$bzip2 = '/usr/bin/bzip2';
|
|
|
|
$data_dir='data';
|
|
|
|
1;
|
|
|
|
sub lock{
|
|
}
|
|
|
|
sub unlock{
|
|
}
|
|
|
|
sub print_time {
|
|
my ($t) = @_;
|
|
my ($minute,$hour,$mday,$mon);
|
|
(undef,$minute,$hour,$mday,$mon,undef) = localtime($t);
|
|
sprintf("%02d/%02d %02d:%02d",$mon+1,$mday,$hour,$minute);
|
|
}
|
|
|
|
#------------------
|
|
# is_today( $time )
|
|
#
|
|
# Takes a Unix time_t and returns true if it represents a time that has
|
|
# the same 'day of year' and 'year' as today. Returns true if no arguments
|
|
# were given.
|
|
#
|
|
sub is_today {
|
|
my ($tt) = @_;
|
|
|
|
my $today = 1;
|
|
|
|
if ( $tt ) {
|
|
my $tt_year = (localtime($tt))[5];
|
|
my $tt_yday = (localtime($tt))[7];
|
|
|
|
my $now_year = (localtime())[5];
|
|
my $now_yday = (localtime())[7];
|
|
|
|
if ( ($tt_year ne $now_year) or ($tt_yday ne $now_yday) ) {
|
|
$today = 0;
|
|
}
|
|
}
|
|
|
|
return $today;
|
|
}
|
|
|
|
#---------------------------------
|
|
# both_are_today( $time1, $time2 )
|
|
#
|
|
# Tests both $time1 and $time2 to see if either are not today. Returns true
|
|
# in that case and false otherwise.
|
|
#
|
|
sub both_are_today {
|
|
my ($t11, $t22) = @_;
|
|
|
|
return 0 if ! is_today($t11);
|
|
return 0 if ! is_today($t22);
|
|
return 1;
|
|
}
|
|
|
|
#---------------------------------
|
|
# get_local_hms( $time, $qualify )
|
|
#
|
|
# Converts a Unix time_t value into a date using format HH:MM. If $qualify
|
|
# is true, it qualifies the date by prepending mm/DD. Returns the resulting
|
|
# string.
|
|
#
|
|
sub get_local_hms {
|
|
my ($t, $need_to_qualify) = @_;
|
|
|
|
my (undef,$minute,$hour,$mday,$mon,undef) = localtime($t);
|
|
my $formatted_date = sprintf("%02d:%02d",$hour,$minute);
|
|
|
|
if ($need_to_qualify) {
|
|
$formatted_date = sprintf("%02d/%02d",$mon+1,$mday) . " " . $formatted_date;
|
|
}
|
|
|
|
return $formatted_date;
|
|
}
|
|
|
|
#--------------------------------------
|
|
# get_time_difference( $time1, $time2 )
|
|
#
|
|
# Calculates a human-readable difference in time between two Unix time_t
|
|
# values. Returns the resulting string.
|
|
#
|
|
sub get_time_difference {
|
|
my ($t11, $t22) = @_;
|
|
|
|
$t11 = 0 if !$t11;
|
|
$t22 = 0 if !$t22;
|
|
|
|
# Flip $t11 and $t22 if $t22 isn't later than $t11.
|
|
if ( $t11 gt $t22 ) {
|
|
my $temp = $t11;
|
|
$t11 = $t22;
|
|
$t22 = $temp;
|
|
}
|
|
|
|
my $time_diff = $t22 - $t11;
|
|
|
|
# @time_slots is an array that we will keep our time difference in.
|
|
# 0 => seconds
|
|
# 1 => minutes
|
|
# 2 => hours
|
|
# 3 => days
|
|
my @time_slots;
|
|
|
|
# Zero out the array.
|
|
for my $ii ( 0 .. 3 ) {
|
|
$time_slots[$ii] = 0;
|
|
}
|
|
|
|
# What's in $time_diff is the difference in seconds. Modulo 60 to get a
|
|
# number of seconds within a minute.
|
|
$time_slots[0] = $time_diff % 60;
|
|
$time_diff = $time_diff / 60;
|
|
|
|
# Now what's left in $time_diff is the difference in minutes. Modulo 60 to
|
|
# get a number of minutes within an hour.
|
|
$time_slots[1] = $time_diff % 60;
|
|
$time_diff = $time_diff / 60;
|
|
|
|
# Now what's left in $time_diff is the difference in hours. Modulo 24 to get
|
|
# a number of hours within a day.
|
|
$time_slots[2] = $time_diff % 24;
|
|
$time_diff = $time_diff / 24;
|
|
|
|
# Now what's left in $time_diff are the number of days. Floor it to get rid
|
|
# of any fractional left from previous maths.
|
|
$time_slots[3] = floor($time_diff);
|
|
|
|
format_time_difference(@time_slots);
|
|
}
|
|
|
|
#--------------------------------------
|
|
# format_time_difference( @time_slots )
|
|
#
|
|
# Called by get_time_difference(), takes its @time_slots arrary and parses it
|
|
# into human-readable text. Returns the resulting string.
|
|
#
|
|
sub format_time_difference {
|
|
my @time_slots = @_;
|
|
|
|
# As above, @time_slots is an array that holds our time difference.
|
|
# 0 => seconds
|
|
# 1 => minutes
|
|
# 2 => hours
|
|
# 3 => days
|
|
|
|
my @formatted;
|
|
|
|
if ( $time_slots[3] eq 1 ) {
|
|
push(@formatted, "1 day");
|
|
} elsif ( $time_slots[3] gt 1 ) {
|
|
push(@formatted, $time_slots[3] . " days");
|
|
}
|
|
|
|
if ( $time_slots[2] eq 1 ) {
|
|
push(@formatted, "1 hour");
|
|
} elsif ( $time_slots[2] gt 1 ) {
|
|
push(@formatted, $time_slots[2] . " hours");
|
|
}
|
|
|
|
if ( $time_slots[1] eq 1 ) {
|
|
push(@formatted, "1 minute");
|
|
} elsif ( $time_slots[1] gt 1 ) {
|
|
push(@formatted, $time_slots[1] . " minutes");
|
|
}
|
|
|
|
# I love that we can track seconds and let the world know, but the unit is
|
|
# too small to be of utility right now. Sticking this code in an always-
|
|
# false conditional in case we want to turn this on in the future.
|
|
if ( 0 ) {
|
|
if ( $time_slots[0] eq 1 ) {
|
|
push(@formatted, "1 second");
|
|
} elsif ( $time_slots[0] gt 1 ) {
|
|
push(@formatted, $time_slots[0] . " seconds");
|
|
}
|
|
}
|
|
|
|
my $formatted_diff;
|
|
if ( scalar(@formatted) gt 0 ) {
|
|
$formatted_diff = join(', ', @formatted);
|
|
} else {
|
|
$formatted_diff = "no time";
|
|
}
|
|
|
|
return $formatted_diff;
|
|
}
|
|
|
|
sub url_encode {
|
|
my ($s) = @_;
|
|
|
|
$s =~ s/\%/\%25/g;
|
|
$s =~ s/\=/\%3d/g;
|
|
$s =~ s/\?/\%3f/g;
|
|
$s =~ s/ /\%20/g;
|
|
$s =~ s/\n/\%0a/g;
|
|
$s =~ s/\r//g;
|
|
$s =~ s/\"/\%22/g;
|
|
$s =~ s/\'/\%27/g;
|
|
$s =~ s/\|/\%7c/g;
|
|
$s =~ s/\&/\%26/g;
|
|
return $s;
|
|
}
|
|
|
|
sub url_decode {
|
|
my ($value) = @_;
|
|
$value =~ tr/+/ /;
|
|
$value =~ s/%([a-fA-F0-9][a-fA-F0-9])/pack("C", hex($1))/eg;
|
|
return $value;
|
|
}
|
|
|
|
|
|
sub value_encode {
|
|
my ($s) = @_;
|
|
$s =~ s@&@&@g;
|
|
$s =~ s@<@<@g;
|
|
$s =~ s@>@>@g;
|
|
$s =~ s@\"@"@g;
|
|
return $s;
|
|
}
|
|
|
|
|
|
BEGIN {
|
|
my $have_loaded_treedata = 0;
|
|
|
|
sub tb_load_treedata {
|
|
my $tree = shift;
|
|
|
|
return if $have_loaded_treedata;
|
|
$have_loaded_treedata = 1;
|
|
require "$tree/treedata.pl" if -r "$tree/treedata.pl";
|
|
}
|
|
}
|
|
|
|
sub tb_load_data {
|
|
$tree = $form{'tree'}; # Testing: $tree = "SeaMonkey";
|
|
|
|
return undef unless $tree;
|
|
|
|
tb_load_treedata($tree);
|
|
|
|
$ignore_builds = {};
|
|
$scrape_builds = {};
|
|
|
|
require "$tree/ignorebuilds.pl" if -r "$tree/ignorebuilds.pl";
|
|
require "$tree/scrapebuilds.pl" if -r "$tree/scrapebuilds.pl";
|
|
|
|
$td = {};
|
|
$td->{name} = $tree;
|
|
$td->{num} = 0;
|
|
$td->{cvs_module} = $cvs_module;
|
|
$td->{cvs_branch} = $cvs_branch;
|
|
$td->{ignore_builds} = $ignore_builds;
|
|
$td->{scrape_builds} = $scrape_builds;
|
|
$cvs_root = '/m/src' if $cvs_root eq '';
|
|
$td->{cvs_root} = $cvs_root;
|
|
|
|
$build_list = load_buildlog($td);
|
|
|
|
get_build_name_index($build_list);
|
|
get_build_time_index($build_list);
|
|
|
|
load_who($td, $who_list);
|
|
|
|
make_build_table($td, $build_list);
|
|
|
|
$td->{scrape} = load_scrape($td);
|
|
$td->{warnings} = load_warnings($td);
|
|
|
|
return $td;
|
|
}
|
|
|
|
sub tb_loadquickparseinfo {
|
|
my ($tree, $build, $times, $includeStatusOfBuilding) = (@_);
|
|
local $_;
|
|
|
|
$maxdate = time;
|
|
require "$tree/ignorebuilds.pl" if -r "$tree/ignorebuilds.pl";
|
|
|
|
my $bw = Backwards->new("$tree/build.dat") or die;
|
|
|
|
my $latest_time = 0;
|
|
my $tooearly = 0;
|
|
while( $_ = $bw->readline ) {
|
|
chop;
|
|
my ($buildtime, $buildname, $buildstatus) = (split /\|/)[1,2,4];
|
|
|
|
if ($includeStatusOfBuilding or
|
|
$buildstatus =~ /^success|busted|testfailed$/) {
|
|
|
|
# Ignore stuff in the future.
|
|
next if $buildtime > $maxdate;
|
|
|
|
$latest_time = $buildtime if $buildtime > $latest_time;
|
|
|
|
# Ignore stuff more than 12 hours old
|
|
if ($buildtime < $latest_time - 12*60*60) {
|
|
# Hack: A build may give a bogus time. To compensate, we will
|
|
# not stop until we hit 20 consecutive lines that are too early.
|
|
# XXX bug 225735: This is the wrong way of doing things. When a
|
|
# flood of backed up mail from one machine comes in, it can easily
|
|
# be more than 20. We should be looking at the mail receipt time
|
|
# to decide when we're done rather than the build time.
|
|
|
|
last if $tooearly++ > 20;
|
|
next;
|
|
}
|
|
$tooearly = 0;
|
|
|
|
next if exists $ignore_builds->{$buildname};
|
|
next if exists $build->{$buildname}
|
|
and $times->{$buildname} >= $buildtime;
|
|
|
|
$build->{$buildname} = $buildstatus;
|
|
$times->{$buildname} = $buildtime;
|
|
}
|
|
}
|
|
}
|
|
|
|
sub tb_last_status {
|
|
my ($build_index) = @_;
|
|
|
|
for (my $tt=0; $tt < $time_count; $tt++) {
|
|
my $br = $build_table->[$tt][$build_index];
|
|
next unless defined $br and $br->{buildstatus};
|
|
next unless $br->{buildstatus} =~ /^(success|busted|testfailed)$/;
|
|
return $br->{buildstatus};
|
|
}
|
|
return 'building';
|
|
}
|
|
|
|
sub tb_check_password {
|
|
if ($form{password} eq '' and defined $cookie_jar{tinderbox_password}) {
|
|
$form{password} = $cookie_jar{tinderbox_password};
|
|
}
|
|
my $correct = '';
|
|
if (open(REAL, '<data/passwd')) {
|
|
$correct = <REAL>;
|
|
close REAL;
|
|
$correct =~ s/\s+$//; # Strip trailing whitespace.
|
|
}
|
|
$form{password} =~ s/\s+$//; # Strip trailing whitespace.
|
|
if ($form{password} ne '') {
|
|
open(TRAPDOOR, "../bonsai/data/trapdoor $form{'password'} |")
|
|
or die "Can't run trapdoor func!";
|
|
my $encoded = <TRAPDOOR>;
|
|
close TRAPDOOR;
|
|
$encoded =~ s/\s+$//; # Strip trailing whitespace.
|
|
if ($encoded eq $correct) {
|
|
if ($form{rememberpassword} ne '') {
|
|
print "Set-Cookie: tinderbox_password=$form{'password'} ;"
|
|
." path=/ ; expires = Sun, 1-Mar-2020 00:00:00 GMT\n";
|
|
}
|
|
return;
|
|
}
|
|
}
|
|
|
|
# Force a return here to test w/o a password.
|
|
# return;
|
|
|
|
require 'header.pl';
|
|
|
|
print "Content-type: text/html\n";
|
|
print "Set-Cookie: tinderbox_password= ; path=/ ; "
|
|
." Expires = Sun, 1-Mar-2020 00:00:00 GMT\n";
|
|
print "\n";
|
|
|
|
EmitHtmlHeader("What's the magic word?",
|
|
"You need to know the magic word to use this page.");
|
|
|
|
if ($form{password} ne '') {
|
|
print "<B>Invalid password; try again.<BR></B>";
|
|
}
|
|
print q(
|
|
<FORM method=post>
|
|
<B>Password:</B>
|
|
<INPUT NAME=password TYPE=password><BR>
|
|
<INPUT NAME=rememberpassword TYPE=checkbox>
|
|
If correct, remember password as a cookie<BR>
|
|
);
|
|
|
|
while (my ($key,$value) = each %form) {
|
|
next if $key eq "password" or $key eq "rememberpassword";
|
|
|
|
my $enc = value_encode($value);
|
|
print "<INPUT TYPE=HIDDEN NAME=$key VALUE=\"$enc\">\n";
|
|
}
|
|
print "<INPUT TYPE=SUBMIT value=Submit></FORM>\n";
|
|
exit;
|
|
}
|
|
|
|
sub tb_find_build_record {
|
|
my ($tree, $logfile) = @_;
|
|
local $_;
|
|
|
|
my $log_entry = '';
|
|
my ($bw) = Backwards->new("$tree/build.dat") or die;
|
|
while( $_ = $bw->readline ) {
|
|
$log_entry = $_ if /$logfile/;
|
|
}
|
|
|
|
chomp($log_entry);
|
|
# Skip the logfile in the parse since it is already known.
|
|
my ($mailtime, $buildtime, $buildname, $errorparser,
|
|
$buildstatus, $binaryurl) = (split /\|/, $log_entry)[0..4,6];
|
|
|
|
$buildrec = {
|
|
mailtime => $mailtime,
|
|
buildtime => $buildtime,
|
|
buildname => $buildname,
|
|
errorparser => $errorparser,
|
|
buildstatus => $buildstatus,
|
|
logfile => $logfile,
|
|
binaryurl => $binaryurl,
|
|
td => undef
|
|
};
|
|
return $buildrec;
|
|
}
|
|
|
|
sub tb_build_static {
|
|
# Build tinderbox static pages
|
|
$ENV{QUERY_STRING}="tree=$tree&static=1";
|
|
$ENV{REQUEST_METHOD}="GET";
|
|
system("./showbuilds.cgi >/dev/null&");
|
|
}
|
|
|
|
# end of public functions
|
|
#============================================================
|
|
|
|
sub load_buildlog {
|
|
my ($treedata) = $_[0];
|
|
|
|
# In general you always want to make "$_" a local
|
|
# if it is used. That way it is restored upon return.
|
|
local $_;
|
|
my $build_list = [];
|
|
|
|
|
|
if (not defined $maxdate) {
|
|
$maxdate = time();
|
|
}
|
|
if (not defined $mindate) {
|
|
$mindate = $maxdate - 24*60*60;
|
|
}
|
|
|
|
my ($bw) = Backwards->new("$treedata->{name}/build.dat") or die;
|
|
|
|
my $tooearly = 0;
|
|
while( $_ = $bw->readline ) {
|
|
chomp;
|
|
my ($mailtime, $buildtime, $buildname,
|
|
$errorparser, $buildstatus, $logfile, $binaryurl) = split /\|/;
|
|
|
|
# Ignore stuff in the future.
|
|
next if $buildtime > $maxdate;
|
|
|
|
# Ignore stuff in the past (but get a 2 hours of extra data)
|
|
if ($buildtime < $mindate - 2*60*60) {
|
|
# Occasionally, a build might show up with a bogus time. So,
|
|
# we won't judge ourselves as having hit the end until we
|
|
# hit a full 20 lines in a row that are too early.
|
|
# XXX bug 225735: This is the wrong way of doing things. When a
|
|
# flood of backed up mail from one machine comes in, it can easily
|
|
# be more than 20. We should be looking at the mail receipt time
|
|
# to decide when we're done rather than the build time.
|
|
last if $tooearly++ > 20;
|
|
|
|
next;
|
|
}
|
|
$tooearly = 0;
|
|
if ($form{noignore} or not $treedata->{ignore_builds}->{$buildname}) {
|
|
my $buildrec = {
|
|
mailtime => $mailtime,
|
|
buildtime => $buildtime,
|
|
buildname => $buildname,
|
|
errorparser => $errorparser,
|
|
buildstatus => $buildstatus,
|
|
logfile => $logfile,
|
|
binaryurl => $binaryurl,
|
|
td => $treedata
|
|
};
|
|
push @{$build_list}, $buildrec;
|
|
}
|
|
}
|
|
return $build_list;
|
|
}
|
|
|
|
# Load data about who checked in when
|
|
# File format: <build_time>|<email_address>
|
|
#
|
|
sub load_who {
|
|
my ($treedata, $who_list) = @_;
|
|
local $_;
|
|
|
|
open(WHOLOG, "<$treedata->{name}/who.dat");
|
|
while (<WHOLOG>) {
|
|
chomp;
|
|
my ($checkin_time, $email) = split /\|/;
|
|
|
|
# Find the time slice where this checkin belongs.
|
|
for (my $ii = $time_count - 1; $ii >= 0; $ii--) {
|
|
if ($checkin_time < $build_time_times->[$ii]) {
|
|
$who_list->[$ii+1]->{$email} = 1;
|
|
last;
|
|
} elsif ($ii == 0) {
|
|
$who_list->[0]->{$email} = 1;
|
|
}
|
|
}
|
|
}
|
|
|
|
# Ignore the last one
|
|
#
|
|
#if ($time_count > 0) {
|
|
# $who_list->[$time_count] = {};
|
|
#}
|
|
}
|
|
|
|
|
|
|
|
# Load data about scrape data.
|
|
# File format: <logfile>|<aaa>|<bbb>|...
|
|
#
|
|
sub load_scrape {
|
|
my $treedata = $_[0];
|
|
local $_;
|
|
|
|
my $scrape = {};
|
|
|
|
open(SCRAPELOG, "<$treedata->{name}/scrape.dat");
|
|
while (<SCRAPELOG>) {
|
|
chomp;
|
|
my @list = split /\|/;
|
|
my $logfile = @list[0];
|
|
shift(@list);
|
|
|
|
$scrape->{$logfile} = [ @list ];
|
|
}
|
|
return $scrape;
|
|
}
|
|
|
|
|
|
# Load data about build warnings
|
|
# File format: <logfile>|<warning_count>
|
|
#
|
|
sub load_warnings {
|
|
my $treedata = $_[0];
|
|
local $_;
|
|
|
|
my $warnings = {};
|
|
|
|
open(WARNINGLOG, "<$treedata->{name}/warnings.dat");
|
|
while (<WARNINGLOG>) {
|
|
chomp;
|
|
my ($logfile, $warning_count) = split /\|/;
|
|
$warnings->{$logfile} = $warning_count;
|
|
}
|
|
return $warnings;
|
|
}
|
|
|
|
sub get_build_name_index {
|
|
my ($build_list) = @_;
|
|
|
|
# Get all the unique build names.
|
|
#
|
|
foreach my $build_record (@{$build_list}) {
|
|
$build_name_index->{$build_record->{buildname}} = 1;
|
|
}
|
|
|
|
my $ii = 0;
|
|
foreach my $name (sort keys %{$build_name_index}) {
|
|
$build_names->[$ii] = $name;
|
|
$build_name_index->{$name} = $ii;
|
|
$ii++;
|
|
}
|
|
$name_count = $#{$build_names} + 1;
|
|
}
|
|
|
|
sub get_build_time_index {
|
|
my ($build_list) = @_;
|
|
|
|
# Get all the unique build names.
|
|
#
|
|
foreach my $br (@{$build_list}) {
|
|
$build_time_index->{$br->{buildtime}} = 1;
|
|
}
|
|
|
|
my $ii = 0;
|
|
foreach my $time (sort {$b <=> $a} keys %{$build_time_index}) {
|
|
$build_time_times->[$ii] = $time;
|
|
$build_time_index->{$time} = $ii;
|
|
$mindate_time_count = $ii if $time >= $mindate;
|
|
$ii++;
|
|
}
|
|
$time_count = $#{$build_time_times} + 1;
|
|
}
|
|
|
|
sub make_build_table {
|
|
my ($treedata, $build_list) = @_;
|
|
my ($ti, $bi, $ti1, $br);
|
|
|
|
# Create the build table
|
|
#
|
|
for (my $ii=0; $ii < $time_count; $ii++){
|
|
$build_table->[$ii] = [];
|
|
}
|
|
|
|
# Populate the build table with build data
|
|
#
|
|
foreach $br (reverse @{$build_list}) {
|
|
$ti = $build_time_index->{$br->{buildtime}};
|
|
$bi = $build_name_index->{$br->{buildname}};
|
|
$build_table->[$ti][$bi] = $br;
|
|
}
|
|
|
|
&load_notes($treedata);
|
|
|
|
for ($bi = $name_count - 1; $bi >= 0; $bi--) {
|
|
for ($ti = $time_count - 1; $ti >= 0; $ti--) {
|
|
if (defined($br = $build_table->[$ti][$bi])
|
|
and not defined($br->{rowspan})) {
|
|
|
|
# If the cell immediately after us is defined, then we
|
|
# can have a previousbuildtime.
|
|
if (defined($br1 = $build_table->[$ti+1][$bi])) {
|
|
$br->{previousbuildtime} = $br1->{buildtime};
|
|
}
|
|
|
|
$ti1 = $ti-1;
|
|
while ($ti1 >= 0 and not defined $build_table->[$ti1][$bi]) {
|
|
$build_table->[$ti1][$bi] = -1;
|
|
$ti1--;
|
|
}
|
|
if ($ti1 > 0 and defined($br1 = $build_table->[$ti1][$bi])) {
|
|
$br->{nextbuildtime} = $br1->{buildtime};
|
|
}
|
|
|
|
$br->{rowspan} = $ti - $ti1;
|
|
unless ($br->{rowspan} == 1) {
|
|
$build_table->[$ti1+1][$bi] = $br;
|
|
$build_table->[$ti][$bi] = -1;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
sub load_notes {
|
|
my $treedata = $_[0];
|
|
|
|
open(NOTES,"<$treedata->{name}/notes.txt")
|
|
or print "<h2>warning: Couldn't open $treedata->{name}/notes.txt </h2>\n";
|
|
while (<NOTES>) {
|
|
chop;
|
|
my ($nbuildtime,$nbuildname,$nwho,$nnow,$nenc_note) = split /\|/;
|
|
my $ti = $build_time_index->{$nbuildtime};
|
|
my $bi = $build_name_index->{$nbuildname};
|
|
|
|
if (defined $ti and defined $bi) {
|
|
$build_table->[$ti][$bi]->{hasnote} = 1;
|
|
unless (defined $build_table->[$ti][$bi]->{noteid}) {
|
|
$build_table->[$ti][$bi]->{noteid} = $#note_array + 1;
|
|
}
|
|
$noteid = $build_table->[$ti][$bi]->{noteid};
|
|
$now_str = &print_time($nnow);
|
|
$note = &url_decode($nenc_note);
|
|
|
|
$note_array[$noteid] = '' unless $note_array[$noteid];
|
|
$note_array[$noteid] = "<pre>\n[<b><a href=mailto:$nwho>"
|
|
."$nwho</a> - $now_str</b>]\n$note\n</pre>"
|
|
.$note_array[$noteid];
|
|
}
|
|
}
|
|
close NOTES;
|
|
}
|
|
|
|
|