gecko-dev/webtools/bugzilla/globals.pl

1520 строки
51 KiB
Perl

# -*- Mode: perl; indent-tabs-mode: nil -*-
#
# The contents of this file are subject to the Mozilla 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/MPL/
#
# 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 Bugzilla Bug Tracking System.
#
# 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): Terry Weissman <terry@mozilla.org>
# Dan Mosedale <dmose@mozilla.org>
# Jacob Steenhagen <jake@bugzilla.org>
# Bradley Baetz <bbaetz@student.usyd.edu.au>
# Christopher Aillon <christopher@aillon.com>
# Joel Peshkin <bugreport@peshkin.net>
# Contains some global variables and routines used throughout bugzilla.
use strict;
use Bugzilla::DB qw(:DEFAULT :deprecated);
use Bugzilla::Constants;
use Bugzilla::Util;
# Bring ChmodDataFile in until this is all moved to the module
use Bugzilla::Config qw(:DEFAULT ChmodDataFile $localconfig $datadir);
# Shut up misguided -w warnings about "used only once". For some reason,
# "use vars" chokes on me when I try it here.
sub globals_pl_sillyness {
my $zz;
$zz = @main::default_column_list;
$zz = $main::defaultqueryname;
$zz = @main::enterable_products;
$zz = %main::keywordsbyname;
$zz = @main::legal_bug_status;
$zz = @main::legal_components;
$zz = @main::legal_keywords;
$zz = @main::legal_opsys;
$zz = @main::legal_platform;
$zz = @main::legal_priority;
$zz = @main::legal_product;
$zz = @main::legal_severity;
$zz = @main::legal_target_milestone;
$zz = @main::legal_versions;
$zz = @main::milestoneurl;
$zz = %main::proddesc;
$zz = @main::prodmaxvotes;
$zz = $main::template;
$zz = $main::userid;
$zz = $main::vars;
}
#
# Here are the --LOCAL-- variables defined in 'localconfig' that we'll use
# here
#
# XXX - Move this to Bugzilla::Config once code which uses these has moved out
# of globals.pl
do $localconfig;
use DBI;
use Date::Format; # For time2str().
use Date::Parse; # For str2time().
use Bugzilla::RelationSet;
# Use standard Perl libraries for cross-platform file/directory manipulation.
use File::Spec;
# Some environment variables are not taint safe
delete @::ENV{'PATH', 'IFS', 'CDPATH', 'ENV', 'BASH_ENV'};
# Cwd.pm in perl 5.6.1 gives a warning if $::ENV{'PATH'} isn't defined
# Set this to '' so that we don't get warnings cluttering the logs on every
# system call
$::ENV{'PATH'} = '';
# Ignore SIGTERM and SIGPIPE - this prevents DB corruption. If the user closes
# their browser window while a script is running, the webserver sends these
# signals, and we don't want to die half way through a write.
$::SIG{TERM} = 'IGNORE';
$::SIG{PIPE} = 'IGNORE';
$::defaultqueryname = "(Default query)"; # This string not exposed in UI
$::unconfirmedstate = "UNCONFIRMED";
# The following subroutine is for debugging purposes only.
# Uncommenting this sub and the $::SIG{__DIE__} trap underneath it will
# cause any fatal errors to result in a call stack trace to help track
# down weird errors.
#sub die_with_dignity {
# use Carp; # for confess()
# my ($err_msg) = @_;
# print $err_msg;
# confess($err_msg);
#}
#$::SIG{__DIE__} = \&die_with_dignity;
@::default_column_list = ("bug_severity", "priority", "rep_platform",
"assigned_to", "bug_status", "resolution",
"short_short_desc");
sub AppendComment {
my ($bugid, $who, $comment, $isprivate, $timestamp, $work_time) = @_;
$work_time ||= 0;
# Use the date/time we were given if possible (allowing calling code
# to synchronize the comment's timestamp with those of other records).
$timestamp = ($timestamp ? SqlQuote($timestamp) : "NOW()");
$comment =~ s/\r\n/\n/g; # Get rid of windows-style line endings.
$comment =~ s/\r/\n/g; # Get rid of mac-style line endings.
# allowing negatives though so people can back out errors in time reporting
if (defined $work_time) {
# regexp verifies one or more digits, optionally followed by a period and
# zero or more digits, OR we have a period followed by one or more digits
if ($work_time !~ /^-?(?:\d+(?:\.\d*)?|\.\d+)$/) {
ThrowUserError("need_numeric_value");
}
} else { $work_time = 0 };
if ($comment =~ /^\s*$/) { # Nothin' but whitespace
return;
}
my $whoid = DBNameToIdAndCheck($who);
my $privacyval = $isprivate ? 1 : 0 ;
SendSQL("INSERT INTO longdescs (bug_id, who, bug_when, thetext, isprivate, work_time) " .
"VALUES($bugid, $whoid, $timestamp, " . SqlQuote($comment) . ", " .
$privacyval . ", " . SqlQuote($work_time) . ")");
SendSQL("UPDATE bugs SET delta_ts = now() WHERE bug_id = $bugid");
}
sub GetFieldID {
my ($f) = (@_);
SendSQL("SELECT fieldid FROM fielddefs WHERE name = " . SqlQuote($f));
my $fieldid = FetchOneColumn();
die "Unknown field id: $f" if !$fieldid;
return $fieldid;
}
# XXXX - this needs to go away
sub GenerateVersionTable {
SendSQL("SELECT versions.value, products.name " .
"FROM versions, products " .
"WHERE products.id = versions.product_id " .
"ORDER BY versions.value");
my @line;
my %varray;
my %carray;
while (@line = FetchSQLData()) {
my ($v,$p1) = (@line);
if (!defined $::versions{$p1}) {
$::versions{$p1} = [];
}
push @{$::versions{$p1}}, $v;
$varray{$v} = 1;
}
SendSQL("SELECT components.name, products.name " .
"FROM components, products " .
"WHERE products.id = components.product_id " .
"ORDER BY components.name");
while (@line = FetchSQLData()) {
my ($c,$p) = (@line);
if (!defined $::components{$p}) {
$::components{$p} = [];
}
my $ref = $::components{$p};
push @$ref, $c;
$carray{$c} = 1;
}
my $dotargetmilestone = 1; # This used to check the param, but there's
# enough code that wants to pretend we're using
# target milestones, even if they don't get
# shown to the user. So we cache all the data
# about them anyway.
my $mpart = $dotargetmilestone ? ", milestoneurl" : "";
SendSQL("select name, description, votesperuser, disallownew$mpart from products ORDER BY name");
while (@line = FetchSQLData()) {
my ($p, $d, $votesperuser, $dis, $u) = (@line);
$::proddesc{$p} = $d;
if (!$dis && scalar($::components{$p})) {
push @::enterable_products, $p;
}
if ($dotargetmilestone) {
$::milestoneurl{$p} = $u;
}
$::prodmaxvotes{$p} = $votesperuser;
}
my $cols = LearnAboutColumns("bugs");
@::log_columns = @{$cols->{"-list-"}};
foreach my $i ("bug_id", "creation_ts", "delta_ts", "lastdiffed") {
my $w = lsearch(\@::log_columns, $i);
if ($w >= 0) {
splice(@::log_columns, $w, 1);
}
}
@::log_columns = (sort(@::log_columns));
@::legal_priority = SplitEnumType($cols->{"priority,type"});
@::legal_severity = SplitEnumType($cols->{"bug_severity,type"});
@::legal_platform = SplitEnumType($cols->{"rep_platform,type"});
@::legal_opsys = SplitEnumType($cols->{"op_sys,type"});
@::legal_bug_status = SplitEnumType($cols->{"bug_status,type"});
@::legal_resolution = SplitEnumType($cols->{"resolution,type"});
# 'settable_resolution' is the list of resolutions that may be set
# directly by hand in the bug form. Start with the list of legal
# resolutions and remove 'MOVED' and 'DUPLICATE' because setting
# bugs to those resolutions requires a special process.
#
@::settable_resolution = @::legal_resolution;
my $w = lsearch(\@::settable_resolution, "DUPLICATE");
if ($w >= 0) {
splice(@::settable_resolution, $w, 1);
}
my $z = lsearch(\@::settable_resolution, "MOVED");
if ($z >= 0) {
splice(@::settable_resolution, $z, 1);
}
my @list = sort { uc($a) cmp uc($b)} keys(%::versions);
@::legal_product = @list;
require File::Temp;
my ($fh, $tmpname) = File::Temp::tempfile("versioncache.XXXXX",
DIR => "$datadir");
print $fh "#\n";
print $fh "# DO NOT EDIT!\n";
print $fh "# This file is automatically generated at least once every\n";
print $fh "# hour by the GenerateVersionTable() sub in globals.pl.\n";
print $fh "# Any changes you make will be overwritten.\n";
print $fh "#\n";
require Data::Dumper;
print $fh (Data::Dumper->Dump([\@::log_columns, \%::versions],
['*::log_columns', '*::versions']));
foreach my $i (@list) {
if (!defined $::components{$i}) {
$::components{$i} = [];
}
}
@::legal_versions = sort {uc($a) cmp uc($b)} keys(%varray);
print $fh (Data::Dumper->Dump([\@::legal_versions, \%::components],
['*::legal_versions', '*::components']));
@::legal_components = sort {uc($a) cmp uc($b)} keys(%carray);
print $fh (Data::Dumper->Dump([\@::legal_components, \@::legal_product,
\@::legal_priority, \@::legal_severity,
\@::legal_platform, \@::legal_opsys,
\@::legal_bug_status, \@::legal_resolution],
['*::legal_components', '*::legal_product',
'*::legal_priority', '*::legal_severity',
'*::legal_platform', '*::legal_opsys',
'*::legal_bug_status', '*::legal_resolution']));
print $fh (Data::Dumper->Dump([\@::settable_resolution, \%::proddesc,
\@::enterable_products, \%::prodmaxvotes],
['*::settable_resolution', '*::proddesc',
'*::enterable_products', '*::prodmaxvotes']));
if ($dotargetmilestone) {
# reading target milestones in from the database - matthew@zeroknowledge.com
SendSQL("SELECT milestones.value, products.name " .
"FROM milestones, products " .
"WHERE products.id = milestones.product_id " .
"ORDER BY milestones.sortkey, milestones.value");
my @line;
my %tmarray;
@::legal_target_milestone = ();
while(@line = FetchSQLData()) {
my ($tm, $pr) = (@line);
if (!defined $::target_milestone{$pr}) {
$::target_milestone{$pr} = [];
}
push @{$::target_milestone{$pr}}, $tm;
if (!exists $tmarray{$tm}) {
$tmarray{$tm} = 1;
push(@::legal_target_milestone, $tm);
}
}
print $fh (Data::Dumper->Dump([\%::target_milestone,
\@::legal_target_milestone,
\%::milestoneurl],
['*::target_milestone',
'*::legal_target_milestone',
'*::milestoneurl']));
}
SendSQL("SELECT id, name FROM keyworddefs ORDER BY name");
while (MoreSQLData()) {
my ($id, $name) = FetchSQLData();
push(@::legal_keywords, $name);
$name = lc($name);
$::keywordsbyname{$name} = $id;
}
print $fh (Data::Dumper->Dump([\@::legal_keywords, \%::keywordsbyname],
['*::legal_keywords', '*::keywordsbyname']));
print $fh "1;\n";
close $fh;
rename $tmpname, "$datadir/versioncache" || die "Can't rename $tmpname to versioncache";
ChmodDataFile("$datadir/versioncache", 0666);
}
sub GetKeywordIdFromName {
my ($name) = (@_);
$name = lc($name);
return $::keywordsbyname{$name};
}
# Returns the modification time of a file.
sub ModTime {
my ($filename) = (@_);
my ($dev,$ino,$mode,$nlink,$uid,$gid,$rdev,$size,
$atime,$mtime,$ctime,$blksize,$blocks)
= stat($filename);
return $mtime;
}
$::VersionTableLoaded = 0;
sub GetVersionTable {
return if $::VersionTableLoaded;
my $mtime = ModTime("$datadir/versioncache");
if (!defined $mtime || $mtime eq "" || !-r "$datadir/versioncache") {
$mtime = 0;
}
if (time() - $mtime > 3600) {
use Bugzilla::Token;
Bugzilla::Token::CleanTokenTable() if Bugzilla->dbwritesallowed;
GenerateVersionTable();
}
require "$datadir/versioncache";
if (!defined %::versions) {
GenerateVersionTable();
do "$datadir/versioncache";
if (!defined %::versions) {
die "Can't generate file $datadir/versioncache";
}
}
$::VersionTableLoaded = 1;
}
# Validates a given username as a new username
# returns 1 if valid, 0 if invalid
sub ValidateNewUser {
my ($username, $old_username) = @_;
if(DBname_to_id($username) != 0) {
return 0;
}
my $sqluname = SqlQuote($username);
# Reject if the new login is part of an email change which is
# still in progress
#
# substring/locate stuff: bug 165221; this used to use regexes, but that
# was unsafe and required weird escaping; using substring to pull out
# the new/old email addresses and locate() to find the delimeter (':')
# is cleaner/safer
SendSQL("SELECT eventdata FROM tokens WHERE tokentype = 'emailold'
AND SUBSTRING(eventdata, 1, (LOCATE(':', eventdata) - 1)) = $sqluname
OR SUBSTRING(eventdata, (LOCATE(':', eventdata) + 1)) = $sqluname");
if (my ($eventdata) = FetchSQLData()) {
# Allow thru owner of token
if($old_username && ($eventdata eq "$old_username:$username")) {
return 1;
}
return 0;
}
return 1;
}
sub InsertNewUser {
my ($username, $realname) = (@_);
# Generate a new random password for the user.
my $password = GenerateRandomPassword();
my $cryptpassword = Crypt($password);
# Insert the new user record into the database.
$username = SqlQuote($username);
$realname = SqlQuote($realname);
$cryptpassword = SqlQuote($cryptpassword);
PushGlobalSQLState();
SendSQL("INSERT INTO profiles (login_name, realname, cryptpassword)
VALUES ($username, $realname, $cryptpassword)");
PopGlobalSQLState();
# Return the password to the calling code so it can be included
# in an email sent to the user.
return $password;
}
sub GenerateRandomPassword {
my ($size) = @_;
# Generated passwords are eight characters long by default.
$size ||= 8;
# The list of characters that can appear in a randomly generated password.
# Note that users can put any character into a password they choose
# themselves.
my @pwchars = (0..9, 'A'..'Z', 'a'..'z', '-', '_', '!', '@', '#', '$',
'%', '^', '*');
# The number of characters in the list.
my $pwcharslen = scalar(@pwchars);
# Generate the password.
my $password = "";
for ( my $i=0 ; $i<$size ; $i++ ) {
$password .= $pwchars[rand($pwcharslen)];
}
# Return the password.
return $password;
}
#
# This function checks if there are any entry groups defined.
# If called with no arguments, it identifies
# entry groups for all products. If called with a product
# id argument, it checks for entry groups associated with
# one particular product.
sub AnyEntryGroups {
my $product_id = shift;
$product_id = 0 unless ($product_id);
return $::CachedAnyEntryGroups{$product_id}
if defined($::CachedAnyEntryGroups{$product_id});
PushGlobalSQLState();
my $query = "SELECT 1 FROM group_control_map WHERE entry != 0";
$query .= " AND product_id = $product_id" if ($product_id);
$query .= " LIMIT 1";
SendSQL($query);
if (MoreSQLData()) {
$::CachedAnyEntryGroups{$product_id} = MoreSQLData();
FetchSQLData();
PopGlobalSQLState();
return $::CachedAnyEntryGroups{$product_id};
} else {
return undef;
}
}
#
# This function checks if there are any default groups defined.
# If so, then groups may have to be changed when bugs move from
# one bug to another.
sub AnyDefaultGroups {
return $::CachedAnyDefaultGroups if defined($::CachedAnyDefaultGroups);
PushGlobalSQLState();
SendSQL("SELECT 1 FROM group_control_map, groups WHERE " .
"groups.id = group_control_map.group_id " .
"AND isactive != 0 AND " .
"(membercontrol = " . CONTROLMAPDEFAULT .
" OR othercontrol = " . CONTROLMAPDEFAULT .
") LIMIT 1");
$::CachedAnyDefaultGroups = MoreSQLData();
FetchSQLData();
PopGlobalSQLState();
return $::CachedAnyDefaultGroups;
}
#
# This function checks if, given a product id, the user can edit
# bugs in this product at all.
sub CanEditProductId {
my ($productid) = @_;
my $query = "SELECT group_id FROM group_control_map " .
"WHERE product_id = $productid " .
"AND canedit != 0 ";
if (defined Bugzilla->user && %{Bugzilla->user->groups}) {
$query .= "AND group_id NOT IN(" .
join(',', values(%{Bugzilla->user->groups})) . ") ";
}
$query .= "LIMIT 1";
PushGlobalSQLState();
SendSQL($query);
my ($result) = FetchSQLData();
PopGlobalSQLState();
return (!defined($result));
}
#
# This function determines if a user can enter bugs in the named
# product.
sub CanEnterProduct {
my ($productname) = @_;
my $query = "SELECT group_id IS NULL " .
"FROM products " .
"LEFT JOIN group_control_map " .
"ON group_control_map.product_id = products.id " .
"AND group_control_map.entry != 0 ";
if (defined Bugzilla->user && %{Bugzilla->user->groups}) {
$query .= "AND group_id NOT IN(" .
join(',', values(%{Bugzilla->user->groups})) . ") ";
}
$query .= "WHERE products.name = " . SqlQuote($productname) . " LIMIT 1";
PushGlobalSQLState();
SendSQL($query);
my ($ret) = FetchSQLData();
PopGlobalSQLState();
return ($ret);
}
#
# This function returns an alphabetical list of product names to which
# the user can enter bugs. If the $by_id parameter is true, also retrieves IDs
# and pushes them onto the list as id, name [, id, name...] for easy slurping
# into a hash by the calling code.
sub GetSelectableProducts {
my ($by_id) = @_;
my $extra_sql = $by_id ? "id, " : "";
my $query = "SELECT $extra_sql name " .
"FROM products " .
"LEFT JOIN group_control_map " .
"ON group_control_map.product_id = products.id ";
if (Param('useentrygroupdefault')) {
$query .= "AND group_control_map.entry != 0 ";
} else {
$query .= "AND group_control_map.membercontrol = " .
CONTROLMAPMANDATORY . " ";
}
if (defined Bugzilla->user && %{Bugzilla->user->groups}) {
$query .= "AND group_id NOT IN(" .
join(',', values(%{Bugzilla->user->groups})) . ") ";
}
$query .= "WHERE group_id IS NULL ORDER BY name";
PushGlobalSQLState();
SendSQL($query);
my @products = ();
push(@products, FetchSQLData()) while MoreSQLData();
PopGlobalSQLState();
return (@products);
}
# GetSelectableProductHash
# returns a hash containing
# legal_products => an enterable product list
# legal_(components|versions|milestones) =>
# the list of components, versions, and milestones of enterable products
# (components|versions|milestones)_by_product
# => a hash of component lists for each enterable product
# Milestones only get returned if the usetargetmilestones parameter is set.
sub GetSelectableProductHash {
# The hash of selectable products and their attributes that gets returned
# at the end of this function.
my $selectables = {};
my %products = GetSelectableProducts(1);
$selectables->{legal_products} = [sort values %products];
# Run queries that retrieve the list of components, versions,
# and target milestones (if used) for the selectable products.
my @tables = qw(components versions);
push(@tables, 'milestones') if Param('usetargetmilestone');
PushGlobalSQLState();
foreach my $table (@tables) {
# Why oh why can't we standardize on these names?!?
my $fld = ($table eq "components" ? "name" : "value");
my $query = "SELECT $fld, product_id FROM $table WHERE product_id IN " .
"(" . join(",", keys %products) . ") ORDER BY $fld";
SendSQL($query);
my %values;
my %values_by_product;
while (MoreSQLData()) {
my ($name, $product_id) = FetchSQLData();
next unless $name;
$values{$name} = 1;
push @{$values_by_product{$products{$product_id}}}, $name;
}
$selectables->{"legal_$table"} = [sort keys %values];
$selectables->{"${table}_by_product"} = \%values_by_product;
}
PopGlobalSQLState();
return $selectables;
}
sub GetFieldDefs {
my $extra = "";
if (!UserInGroup(Param('timetrackinggroup'))) {
$extra = "WHERE name NOT IN ('estimated time', 'remaining_time', " .
"'work_time', 'percentage_complete')";
}
my @fields;
PushGlobalSQLState();
SendSQL("SELECT name, description FROM fielddefs $extra ORDER BY sortkey");
while (MoreSQLData()) {
my ($name, $description) = FetchSQLData();
push(@fields, { name => $name, description => $description });
}
PopGlobalSQLState();
return(@fields);
}
sub CanSeeBug {
my ($id, $userid) = @_;
# Query the database for the bug, retrieving a boolean value that
# represents whether or not the user is authorized to access the bug.
# if no groups are found --> user is permitted to access
# if no user is found for any group --> user is not permitted to access
my $query = "SELECT bugs.bug_id, reporter, assigned_to, qa_contact," .
" reporter_accessible, cclist_accessible," .
" cc.who IS NOT NULL," .
" COUNT(DISTINCT(bug_group_map.group_id)) as cntbugingroups," .
" COUNT(DISTINCT(user_group_map.group_id)) as cntuseringroups" .
" FROM bugs" .
" LEFT JOIN cc ON bugs.bug_id = cc.bug_id" .
" AND cc.who = $userid" .
" LEFT JOIN bug_group_map ON bugs.bug_id = bug_group_map.bug_id" .
" LEFT JOIN user_group_map ON" .
" user_group_map.group_id = bug_group_map.group_id" .
" AND user_group_map.isbless = 0" .
" AND user_group_map.user_id = $userid" .
" WHERE bugs.bug_id = $id GROUP BY bugs.bug_id";
PushGlobalSQLState();
SendSQL($query);
my ($found_id, $reporter, $assigned_to, $qa_contact,
$rep_access, $cc_access,
$found_cc, $found_groups, $found_members)
= FetchSQLData();
PopGlobalSQLState();
return (
($found_groups == 0)
|| (($userid > 0) &&
(
($assigned_to == $userid)
|| (Param('useqacontact') && $qa_contact == $userid)
|| (($reporter == $userid) && $rep_access)
|| ($found_cc && $cc_access)
|| ($found_groups == $found_members)
))
);
}
sub ValidatePassword {
# Determines whether or not a password is valid (i.e. meets Bugzilla's
# requirements for length and content).
# If a second password is passed in, this function also verifies that
# the two passwords match.
my ($password, $matchpassword) = @_;
if (length($password) < 3) {
ThrowUserError("password_too_short");
} elsif (length($password) > 16) {
ThrowUserError("password_too_long");
} elsif ((defined $matchpassword) && ($password ne $matchpassword)) {
ThrowUserError("passwords_dont_match");
}
}
sub Crypt {
# Crypts a password, generating a random salt to do it.
# Random salts are generated because the alternative is usually
# to use the first two characters of the password itself, and since
# the salt appears in plaintext at the beginning of the crypted
# password string this has the effect of revealing the first two
# characters of the password to anyone who views the crypted version.
my ($password) = @_;
# The list of characters that can appear in a salt. Salts and hashes
# are both encoded as a sequence of characters from a set containing
# 64 characters, each one of which represents 6 bits of the salt/hash.
# The encoding is similar to BASE64, the difference being that the
# BASE64 plus sign (+) is replaced with a forward slash (/).
my @saltchars = (0..9, 'A'..'Z', 'a'..'z', '.', '/');
# Generate the salt. We use an 8 character (48 bit) salt for maximum
# security on systems whose crypt uses MD5. Systems with older
# versions of crypt will just use the first two characters of the salt.
my $salt = '';
for ( my $i=0 ; $i < 8 ; ++$i ) {
$salt .= $saltchars[rand(64)];
}
# Crypt the password.
my $cryptedpassword = crypt($password, $salt);
# Return the crypted password.
return $cryptedpassword;
}
sub DBID_to_real_or_loginname {
my ($id) = (@_);
PushGlobalSQLState();
SendSQL("SELECT login_name,realname FROM profiles WHERE userid = $id");
my ($l, $r) = FetchSQLData();
PopGlobalSQLState();
if (!defined $r || $r eq "") {
return $l;
} else {
return "$l ($r)";
}
}
sub DBID_to_name {
my ($id) = (@_);
# $id should always be a positive integer
if ($id =~ m/^([1-9][0-9]*)$/) {
$id = $1;
} else {
$::cachedNameArray{$id} = "__UNKNOWN__";
}
if (!defined $::cachedNameArray{$id}) {
PushGlobalSQLState();
SendSQL("select login_name from profiles where userid = $id");
my $r = FetchOneColumn();
PopGlobalSQLState();
if (!defined $r || $r eq "") {
$r = "__UNKNOWN__";
}
$::cachedNameArray{$id} = $r;
}
return $::cachedNameArray{$id};
}
sub DBname_to_id {
my ($name) = (@_);
PushGlobalSQLState();
SendSQL("select userid from profiles where login_name = @{[SqlQuote($name)]}");
my $r = FetchOneColumn();
PopGlobalSQLState();
# $r should be a positive integer, this makes Taint mode happy
if (defined $r && $r =~ m/^([1-9][0-9]*)$/) {
return $1;
} else {
return 0;
}
}
sub DBNameToIdAndCheck {
my ($name) = (@_);
my $result = DBname_to_id($name);
if ($result > 0) {
return $result;
}
ThrowUserError("invalid_username",
{ name => $name });
}
sub get_product_id {
my ($prod) = @_;
PushGlobalSQLState();
SendSQL("SELECT id FROM products WHERE name = " . SqlQuote($prod));
my ($prod_id) = FetchSQLData();
PopGlobalSQLState();
return $prod_id;
}
sub get_product_name {
my ($prod_id) = @_;
die "non-numeric prod_id '$prod_id' passed to get_product_name"
unless ($prod_id =~ /^\d+$/);
PushGlobalSQLState();
SendSQL("SELECT name FROM products WHERE id = $prod_id");
my ($prod) = FetchSQLData();
PopGlobalSQLState();
return $prod;
}
sub get_component_id {
my ($prod_id, $comp) = @_;
return undef unless ($prod_id && ($prod_id =~ /^\d+$/));
PushGlobalSQLState();
SendSQL("SELECT id FROM components " .
"WHERE product_id = $prod_id AND name = " . SqlQuote($comp));
my ($comp_id) = FetchSQLData();
PopGlobalSQLState();
return $comp_id;
}
sub get_component_name {
my ($comp_id) = @_;
die "non-numeric comp_id '$comp_id' passed to get_component_name"
unless ($comp_id =~ /^\d+$/);
PushGlobalSQLState();
SendSQL("SELECT name FROM components WHERE id = $comp_id");
my ($comp) = FetchSQLData();
PopGlobalSQLState();
return $comp;
}
# This routine quoteUrls contains inspirations from the HTML::FromText CPAN
# module by Gareth Rees <garethr@cre.canon.co.uk>. It has been heavily hacked,
# all that is really recognizable from the original is bits of the regular
# expressions.
# This has been rewritten to be faster, mainly by substituting 'as we go'.
# If you want to modify this routine, read the comments carefully
sub quoteUrls {
my ($text) = (@_);
return $text unless $text;
# We use /g for speed, but uris can have other things inside them
# (http://foo/bug#3 for example). Filtering that out filters valid
# bug refs out, so we have to do replacements.
# mailto can't contain space or #, so we don't have to bother for that
# Do this by escaping \0 to \1\0, and replacing matches with \0\0$count\0\0
# \0 is used because its unliklely to occur in the text, so the cost of
# doing this should be very small
# Also, \0 won't appear in the value_quote'd bug title, so we don't have
# to worry about bogus substitutions from there
# escape the 2nd escape char we're using
my $chr1 = chr(1);
$text =~ s/\0/$chr1\0/g;
# However, note that adding the title (for buglinks) can affect things
# In particular, attachment matches go before bug titles, so that titles
# with 'attachment 1' don't double match.
# Dupe checks go afterwards, because that uses ^ and \Z, which won't occur
# if it was subsituted as a bug title (since that always involve leading
# and trailing text)
# Because of entities, its easier (and quicker) to do this before escaping
my @things;
my $count = 0;
my $tmp;
# non-mailto protocols
my $protocol_re = qr/(afs|cid|ftp|gopher|http|https|irc|mid|news|nntp|prospero|telnet|view-source|wais)/i;
$text =~ s~\b(${protocol_re}: # The protocol:
[^\s<>\"]+ # Any non-whitespace
[\w\/]) # so that we end in \w or /
~($tmp = html_quote($1)) &&
($things[$count++] = "<a href=\"$tmp\">$tmp</a>") &&
("\0\0" . ($count-1) . "\0\0")
~egox;
# We have to quote now, otherwise our html is itsself escaped
# THIS MEANS THAT A LITERAL ", <, >, ' MUST BE ESCAPED FOR A MATCH
$text = html_quote($text);
# mailto:
# Use |<nothing> so that $1 is defined regardless
$text =~ s~\b(mailto:|)?([\w\.\-\+\=]+\@[\w\-]+(?:\.[\w\-]+)+)\b
~<a href=\"mailto:$2\">$1$2</a>~igx;
# attachment links - handle both cases separatly for simplicity
$text =~ s~((?:^Created\ an\ |\b)attachment\s*\(id=(\d+)\))
~GetAttachmentLink($2, $1)
~egmx;
$text =~ s~\b(attachment\s*\#?\s*(\d+))
~GetAttachmentLink($2, $1)
~egmxi;
# This handles bug a, comment b type stuff. Because we're using /g
# we have to do this in one pattern, and so this is semi-messy.
# Also, we can't use $bug_re?$comment_re? because that will match the
# empty string
my $bug_re = qr/bug\s*\#?\s*(\d+)/i;
my $comment_re = qr/comment\s*\#?\s*(\d+)/i;
$text =~ s~\b($bug_re(?:\s*,?\s*$comment_re)?|$comment_re)
~ # We have several choices. $1 here is the link, and $2-4 are set
# depending on which part matched
(defined($2) ? GetBugLink($2,$1,$3) :
"<a href=\"#c$4\">$1</a>")
~egox;
# Duplicate markers
$text =~ s~(?<=^\*\*\*\ This\ bug\ has\ been\ marked\ as\ a\ duplicate\ of\ )
(\d+)
(?=\ \*\*\*\Z)
~GetBugLink($1, $1)
~egmx;
# Now remove the encoding hacks
$text =~ s/\0\0(\d+)\0\0/$things[$1]/eg;
$text =~ s/$chr1\0/\0/g;
return $text;
}
# GetAttachmentLink creates a link to an attachment,
# including its title.
sub GetAttachmentLink {
my ($attachid, $link_text) = @_;
detaint_natural($attachid) ||
die "GetAttachmentLink() called with non-integer attachment number";
# If we've run GetAttachmentLink() for this attachment before,
# %::attachlink will contain an anonymous array ref of relevant
# values. If not, we need to get the information from the database.
if (! defined $::attachlink{$attachid}) {
# Make sure any unfetched data from a currently running query
# is saved off rather than overwritten
PushGlobalSQLState();
SendSQL("SELECT bug_id, isobsolete, description
FROM attachments WHERE attach_id = $attachid");
if (MoreSQLData()) {
my ($bugid, $isobsolete, $desc) = FetchSQLData();
my $title = "";
my $className = "";
if (CanSeeBug($bugid, $::userid)) {
$title = $desc;
}
if ($isobsolete) {
$className = "bz_obsolete";
}
$::attachlink{$attachid} = [value_quote($title), $className];
}
else {
# Even if there's nothing in the database, we want to save a blank
# anonymous array in the %::attachlink hash so the query doesn't get
# run again next time we're called for this attachment number.
$::attachlink{$attachid} = [];
}
# All done with this sidetrip
PopGlobalSQLState();
}
# Now that we know we've got all the information we're gonna get, let's
# return the link (which is the whole reason we were called :)
my ($title, $className) = @{$::attachlink{$attachid}};
# $title will be undefined if the bug didn't exist in the database.
if (defined $title) {
my $linkval = "attachment.cgi?id=$attachid&amp;action=view";
return qq{<a href="$linkval" class="$className" title="$title">$link_text</a>};
}
else {
return qq{$link_text};
}
}
# GetBugLink creates a link to a bug, including its title.
# It takes either two or three parameters:
# - The bug number
# - The link text, to place between the <a>..</a>
# - An optional comment number, for linking to a particular
# comment in the bug
sub GetBugLink {
my ($bug_num, $link_text, $comment_num) = @_;
$bug_num || return "&lt;missing bug number&gt;";
detaint_natural($bug_num) || return "&lt;invalid bug number&gt;";
# If we've run GetBugLink() for this bug number before, %::buglink
# will contain an anonymous array ref of relevent values, if not
# we need to get the information from the database.
if (! defined $::buglink{$bug_num}) {
# Make sure any unfetched data from a currently running query
# is saved off rather than overwritten
PushGlobalSQLState();
SendSQL("SELECT bugs.bug_status, resolution, short_desc " .
"FROM bugs WHERE bugs.bug_id = $bug_num");
# If the bug exists, save its data off for use later in the sub
if (MoreSQLData()) {
my ($bug_state, $bug_res, $bug_desc) = FetchSQLData();
# Initialize these variables to be "" so that we don't get warnings
# if we don't change them below (which is highly likely).
my ($pre, $title, $post) = ("", "", "");
$title = $bug_state;
if ($bug_state eq $::unconfirmedstate) {
$pre = "<i>";
$post = "</i>";
}
elsif (! IsOpenedState($bug_state)) {
$pre = '<span class="bz_closed">';
$title .= " $bug_res";
$post = '</span>';
}
if (CanSeeBug($bug_num, $::userid)) {
$title .= " - $bug_desc";
}
$::buglink{$bug_num} = [$pre, value_quote($title), $post];
}
else {
# Even if there's nothing in the database, we want to save a blank
# anonymous array in the %::buglink hash so the query doesn't get
# run again next time we're called for this bug number.
$::buglink{$bug_num} = [];
}
# All done with this sidetrip
PopGlobalSQLState();
}
# Now that we know we've got all the information we're gonna get, let's
# return the link (which is the whole reason we were called :)
my ($pre, $title, $post) = @{$::buglink{$bug_num}};
# $title will be undefined if the bug didn't exist in the database.
if (defined $title) {
my $linkval = "show_bug.cgi?id=$bug_num";
if (defined $comment_num) {
$linkval .= "#c$comment_num";
}
return qq{$pre<a href="$linkval" title="$title">$link_text</a>$post};
}
else {
return qq{$link_text};
}
}
# CountOpenDependencies counts the number of open dependent bugs for a
# list of bugs and returns a list of bug_id's and their dependency count
# It takes one parameter:
# - A list of bug numbers whose dependencies are to be checked
sub CountOpenDependencies {
my (@bug_list) = @_;
my @dependencies;
# Make sure any unfetched data from a currently running query
# is saved off rather than overwritten
PushGlobalSQLState();
SendSQL("SELECT blocked, count(bug_status) " .
"FROM bugs, dependencies " .
"WHERE blocked IN (" . (join "," , @bug_list) . ") " .
"AND bug_id = dependson " .
"AND bug_status IN ('" . (join "','", OpenStates()) . "') " .
"GROUP BY blocked ");
while (MoreSQLData()) {
my ($bug_id, $dependencies) = FetchSQLData();
push(@dependencies, { bug_id => $bug_id,
dependencies => $dependencies });
}
# All done with this sidetrip
PopGlobalSQLState();
return @dependencies;
}
sub GetLongDescriptionAsText {
my ($id, $start, $end) = (@_);
my $result = "";
my $count = 0;
my $anyprivate = 0;
my ($query) = ("SELECT profiles.login_name, DATE_FORMAT(longdescs.bug_when,'%Y.%m.%d %H:%i'), " .
" longdescs.thetext, longdescs.isprivate " .
"FROM longdescs, profiles " .
"WHERE profiles.userid = longdescs.who " .
"AND longdescs.bug_id = $id ");
if ($start && $start =~ /[1-9]/) {
# If the start is all zeros, then don't do this (because we want to
# not emit a leading "Additional Comments" line in that case.)
$query .= "AND longdescs.bug_when > '$start'";
$count = 1;
}
if ($end) {
$query .= "AND longdescs.bug_when <= '$end'";
}
$query .= "ORDER BY longdescs.bug_when";
SendSQL($query);
while (MoreSQLData()) {
my ($who, $when, $text, $isprivate, $work_time) = (FetchSQLData());
if ($count) {
$result .= "\n\n------- Additional Comments From $who".Param('emailsuffix')." ".
Bugzilla::Util::format_time($when) . " -------\n";
}
if (($isprivate > 0) && Param("insidergroup")) {
$anyprivate = 1;
}
$result .= $text;
$count++;
}
return ($result, $anyprivate);
}
sub GetComments {
my ($id) = (@_);
my @comments;
SendSQL("SELECT profiles.realname, profiles.login_name,
date_format(longdescs.bug_when,'%Y.%m.%d %H:%i'),
longdescs.thetext, longdescs.work_time,
isprivate,
date_format(longdescs.bug_when,'%Y%m%d%H%i%s')
FROM longdescs, profiles
WHERE profiles.userid = longdescs.who
AND longdescs.bug_id = $id
ORDER BY longdescs.bug_when");
while (MoreSQLData()) {
my %comment;
($comment{'name'}, $comment{'email'}, $comment{'time'},
$comment{'body'}, $comment{'work_time'},
$comment{'isprivate'}, $comment{'when'}) = FetchSQLData();
$comment{'email'} .= Param('emailsuffix');
$comment{'name'} = $comment{'name'} || $comment{'email'};
push (@comments, \%comment);
}
return \@comments;
}
# Fills in a hashtable with info about the columns for the given table in the
# database. The hashtable has the following entries:
# -list- the list of column names
# <name>,type the type for the given name
sub LearnAboutColumns {
my ($table) = (@_);
my %a;
SendSQL("show columns from $table");
my @list = ();
my @row;
while (@row = FetchSQLData()) {
my ($name,$type) = (@row);
$a{"$name,type"} = $type;
push @list, $name;
}
$a{"-list-"} = \@list;
return \%a;
}
# If the above returned a enum type, take that type and parse it into the
# list of values. Assumes that enums don't ever contain an apostrophe!
sub SplitEnumType {
my ($str) = (@_);
my @result = ();
if ($str =~ /^enum\((.*)\)$/) {
my $guts = $1 . ",";
while ($guts =~ /^\'([^\']*)\',(.*)$/) {
push @result, $1;
$guts = $2;
}
}
return @result;
}
sub UserInGroup {
if ($_[1]) {
die "UserInGroup no longer takes a second parameter.";
}
return defined Bugzilla->user && defined Bugzilla->user->groups->{$_[0]};
}
sub UserCanBlessGroup {
my ($groupname) = (@_);
PushGlobalSQLState();
# check if user explicitly can bless group
SendSQL("SELECT groups.id FROM groups, user_group_map
WHERE groups.id = user_group_map.group_id
AND user_group_map.user_id = $::userid
AND isbless = 1
AND groups.name = " . SqlQuote($groupname));
my $result = FetchOneColumn();
PopGlobalSQLState();
if ($result) {
return 1;
}
PushGlobalSQLState();
# check if user is a member of a group that can bless this group
# this group does not count
SendSQL("SELECT groups.id FROM groups, user_group_map,
group_group_map
WHERE groups.id = grantor_id
AND user_group_map.user_id = $::userid
AND user_group_map.isbless = 0
AND group_group_map.isbless = 1
AND user_group_map.group_id = member_id
AND groups.name = " . SqlQuote($groupname));
$result = FetchOneColumn();
PopGlobalSQLState();
return $result;
}
sub BugInGroup {
my ($bugid, $groupname) = (@_);
PushGlobalSQLState();
SendSQL("SELECT bug_group_map.bug_id != 0 FROM bug_group_map, groups
WHERE bug_group_map.bug_id = $bugid
AND bug_group_map.group_id = groups.id
AND groups.name = " . SqlQuote($groupname));
my $bugingroup = FetchOneColumn();
PopGlobalSQLState();
return $bugingroup;
}
sub BugInGroupId {
my ($bugid, $groupid) = (@_);
PushGlobalSQLState();
SendSQL("SELECT bug_id != 0 FROM bug_group_map
WHERE bug_id = $bugid
AND group_id = $groupid");
my $bugingroup = FetchOneColumn();
PopGlobalSQLState();
return $bugingroup;
}
sub GroupExists {
my ($groupname) = (@_);
PushGlobalSQLState();
SendSQL("SELECT id FROM groups WHERE name=" . SqlQuote($groupname));
my $id = FetchOneColumn();
PopGlobalSQLState();
return $id;
}
sub GroupNameToId {
my ($groupname) = (@_);
PushGlobalSQLState();
SendSQL("SELECT id FROM groups WHERE name=" . SqlQuote($groupname));
my $id = FetchOneColumn();
PopGlobalSQLState();
return $id;
}
sub GroupIdToName {
my ($groupid) = (@_);
PushGlobalSQLState();
SendSQL("SELECT name FROM groups WHERE id = $groupid");
my $name = FetchOneColumn();
PopGlobalSQLState();
return $name;
}
# Determines whether or not a group is active by checking
# the "isactive" column for the group in the "groups" table.
# Note: This function selects groups by id rather than by name.
sub GroupIsActive {
my ($groupid) = (@_);
$groupid ||= 0;
PushGlobalSQLState();
SendSQL("SELECT isactive FROM groups WHERE id=$groupid");
my $isactive = FetchOneColumn();
PopGlobalSQLState();
return $isactive;
}
# Determines if the given bug_status string represents an "Opened" bug. This
# routine ought to be parameterizable somehow, as people tend to introduce
# new states into Bugzilla.
sub IsOpenedState {
my ($state) = (@_);
if (grep($_ eq $state, OpenStates())) {
return 1;
}
return 0;
}
# This sub will return an array containing any status that
# is considered an open bug.
sub OpenStates {
return ('NEW', 'REOPENED', 'ASSIGNED', $::unconfirmedstate);
}
sub RemoveVotes {
my ($id, $who, $reason) = (@_);
my $whopart = "";
if ($who) {
$whopart = " AND votes.who = $who";
}
SendSQL("SELECT profiles.login_name, profiles.userid, votes.vote_count, " .
"products.votesperuser, products.maxvotesperbug " .
"FROM profiles " .
"LEFT JOIN votes ON profiles.userid = votes.who " .
"LEFT JOIN bugs USING(bug_id) " .
"LEFT JOIN products ON products.id = bugs.product_id " .
"WHERE votes.bug_id = $id " .
$whopart);
my @list;
while (MoreSQLData()) {
my ($name, $userid, $oldvotes, $votesperuser, $maxvotesperbug) = (FetchSQLData());
push(@list, [$name, $userid, $oldvotes, $votesperuser, $maxvotesperbug]);
}
if (0 < @list) {
foreach my $ref (@list) {
my ($name, $userid, $oldvotes, $votesperuser, $maxvotesperbug) = (@$ref);
my $s;
$maxvotesperbug = $votesperuser if ($votesperuser < $maxvotesperbug);
# If this product allows voting and the user's votes are in
# the acceptable range, then don't do anything.
next if $votesperuser && $oldvotes <= $maxvotesperbug;
# If the user has more votes on this bug than this product
# allows, then reduce the number of votes so it fits
my $newvotes = $votesperuser ? $maxvotesperbug : 0;
my $removedvotes = $oldvotes - $newvotes;
$s = $oldvotes == 1 ? "" : "s";
my $oldvotestext = "You had $oldvotes vote$s on this bug.";
$s = $removedvotes == 1 ? "" : "s";
my $removedvotestext = "You had $removedvotes vote$s removed from this bug.";
my $newvotestext;
if ($newvotes) {
SendSQL("UPDATE votes SET vote_count = $newvotes " .
"WHERE bug_id = $id AND who = $userid");
$s = $newvotes == 1 ? "" : "s";
$newvotestext = "You still have $newvotes vote$s on this bug."
} else {
SendSQL("DELETE FROM votes WHERE bug_id = $id AND who = $userid");
$newvotestext = "You have no more votes remaining on this bug.";
}
# Notice that we did not make sure that the user fit within the $votesperuser
# range. This is considered to be an acceptable alternative to losing votes
# during product moves. Then next time the user attempts to change their votes,
# they will be forced to fit within the $votesperuser limit.
# Now lets send the e-mail to alert the user to the fact that their votes have
# been reduced or removed.
my $sendmailparm = '-ODeliveryMode=deferred';
if (Param('sendmailnow')) {
$sendmailparm = '';
}
if (open(SENDMAIL, "|/usr/lib/sendmail $sendmailparm -t -i")) {
my %substs;
$substs{"to"} = $name;
$substs{"bugid"} = $id;
$substs{"reason"} = $reason;
$substs{"votesremoved"} = $removedvotes;
$substs{"votesold"} = $oldvotes;
$substs{"votesnew"} = $newvotes;
$substs{"votesremovedtext"} = $removedvotestext;
$substs{"votesoldtext"} = $oldvotestext;
$substs{"votesnewtext"} = $newvotestext;
$substs{"count"} = $removedvotes . "\n " . $newvotestext;
my $msg = PerformSubsts(Param("voteremovedmail"),
\%substs);
print SENDMAIL $msg;
close SENDMAIL;
}
}
SendSQL("SELECT SUM(vote_count) FROM votes WHERE bug_id = $id");
my $v = FetchOneColumn();
$v ||= 0;
SendSQL("UPDATE bugs SET votes = $v, delta_ts = delta_ts " .
"WHERE bug_id = $id");
}
}
# Take two comma or space separated strings and return what
# values were removed from or added to the new one.
sub DiffStrings {
my ($oldstr, $newstr) = @_;
# Split the old and new strings into arrays containing their values.
$oldstr =~ s/[\s,]+/ /g;
$newstr =~ s/[\s,]+/ /g;
my @old = split(" ", $oldstr);
my @new = split(" ", $newstr);
my (@remove, @add) = ();
# Find values that were removed
foreach my $value (@old) {
push (@remove, $value) if !grep($_ eq $value, @new);
}
# Find values that were added
foreach my $value (@new) {
push (@add, $value) if !grep($_ eq $value, @old);
}
my $removed = join (", ", @remove);
my $added = join (", ", @add);
return ($removed, $added);
}
sub PerformSubsts {
my ($str, $substs) = (@_);
$str =~ s/%([a-z]*)%/(defined $substs->{$1} ? $substs->{$1} : Param($1))/eg;
return $str;
}
sub FormatTimeUnit {
# Returns a number with 2 digit precision, unless the last digit is a 0
# then it returns only 1 digit precision
my ($time) = (@_);
my $newtime = sprintf("%.2f", $time);
if ($newtime =~ /0\Z/) {
$newtime = sprintf("%.1f", $time);
}
return $newtime;
}
###############################################################################
# Constructs a format object from URL parameters. You most commonly call it
# like this:
# my $format = GetFormat("foo/bar", scalar($cgi->param('format')),
# scalar($cgi->param('ctype')));
sub GetFormat {
my ($template, $format, $ctype) = @_;
$ctype ||= "html";
$format ||= "";
# Security - allow letters and a hyphen only
$ctype =~ s/[^a-zA-Z\-]//g;
$format =~ s/[^a-zA-Z\-]//g;
trick_taint($ctype);
trick_taint($format);
$template .= ($format ? "-$format" : "");
$template .= ".$ctype.tmpl";
# Now check that the template actually exists. We only want to check
# if the template exists; any other errors (eg parse errors) will
# end up being detected later.
eval {
Bugzilla->template->context->template($template);
};
# This parsing may seem fragile, but its OK:
# http://lists.template-toolkit.org/pipermail/templates/2003-March/004370.html
# Even if it is wrong, any sort of error is going to cause a failure
# eventually, so the only issue would be an incorrect error message
if ($@ && $@->info =~ /: not found$/) {
ThrowUserError("format_not_found", { 'format' => $format,
'ctype' => $ctype,
});
}
# Else, just return the info
return
{
'template' => $template ,
'extension' => $ctype ,
'ctype' => Bugzilla::Constants::contenttypes->{$ctype} ,
};
}
############# Live code below here (that is, not subroutine defs) #############
use Bugzilla;
$::template = Bugzilla->template();
$::vars = {};
1;