2 # Copyright (C) 2014-2021 all contributors <meta@public-inbox.org>
3 # License: AGPL-3.0+ <https://www.gnu.org/licenses/agpl-3.0.txt>
6 use Getopt::Long qw/:config gnu_getopt no_ignore_case auto_abbrev/;
7 use Fcntl qw(:DEFAULT);
8 my $help = <<EOF; # the following should fit w/o scrolling in 80x24 term:
9 usage: public-inbox-init NAME INBOX_DIR HTTP_URL ADDRESS [ADDRESS..]
11 Initialize a public-inbox
15 NAME the name of the inbox
16 INBOX_DIR pathname the inbox
17 HTTP_URL HTTP (or HTTPS) URL
18 ADDRESS email address(es), may be specified multiple times
22 -V2 use scalable public-inbox-v2-format(5)
23 -L LEVEL index level `basic', `medium', or `full' (default: full)
24 --ng NEWSGROUP set NNTP newsgroup name
25 -c KEY=VALUE set additional config option(s)
26 --skip-artnum=NUM NNTP article numbers to skip
27 --skip-epoch=NUM epochs to skip (-V2 only)
28 -j JOBS number of indexing jobs (-V2 only), (default: 4)
30 See public-inbox-init(1) man page for full documentation.
33 require PublicInbox::Admin;
34 PublicInbox::Admin::require_or_die('-base');
36 my ($version, $indexlevel, $skip_epoch, $skip_artnum, $jobs, $show_help);
41 'V|version=i' => \$version,
42 'L|index-level|indexlevel=s' => \$indexlevel,
43 'S|skip|skip-epoch=i' => \$skip_epoch,
44 'skip-artnum=i' => \$skip_artnum,
46 'ng|newsgroup=s' => \$ng,
47 'skip-docdata' => \$skip_docdata,
48 'help|h' => \$show_help,
55 GetOptions(%opts) or $usage_cb->();
56 if ($show_help) { print $help; exit 0 };
57 my $name = shift @ARGV or $usage_cb->();
58 my $inboxdir = shift @ARGV or $usage_cb->();
59 my $http_url = shift @ARGV or $usage_cb->();
60 my (@address) = @ARGV;
61 @address or $usage_cb->();
64 my ($k, $v) = split(/=/, $_, 2);
65 defined($v) or die "Usage: -c KEY=VALUE\n";
66 $k =~ /\A[a-z]+\z/i or die "$k contains invalid characters\n";
68 if ($k eq 'newsgroup') {
69 die "newsgroup already set ($ng)\n" if $ng ne '';
72 } elsif ($k eq 'address') {
73 push @address, $v; # for conflict checking
75 } elsif ($k =~ /\A(?:inboxdir|mainrepo)\z/) {
76 die "$k not allowed via -c $_\n"
77 } elsif ($k eq 'indexlevel') {
78 defined($indexlevel) and
79 die "indexlevel already set ($indexlevel)\n";
87 PublicInbox::Admin::indexlevel_ok_or_die($indexlevel) if defined $indexlevel;
89 $ng =~ m![^A-Za-z0-9/_\.\-\~\@\+\=:]! and
90 die "--newsgroup `$ng' is not valid\n";
91 ($ng =~ m!\A\.! || $ng =~ m!\.\z!) and
92 die "--newsgroup `$ng' must not start or end with `.'\n";
94 require PublicInbox::Config;
95 my $pi_config = PublicInbox::Config->default_file;
96 require File::Basename;
97 my $dir = File::Basename::dirname($pi_config);
99 File::Path::mkpath($dir); # will croak on fatal errors
101 # first, we grab a flock to prevent simultaneous public-inbox-init
102 # processes from trampling over each other, or exiting with 255 on
103 # O_EXCL failure below. This gets unlocked automatically on exit:
104 require PublicInbox::Lock;
105 my $lock_obj = { lock_path => "$pi_config.flock" };
106 PublicInbox::Lock::lock_acquire($lock_obj);
108 # git-config will operate on this (and rename on success):
110 my $fh = File::Temp->new(TEMPLATE => 'pi-init-XXXX', DIR => $dir);
112 # Now, we grab another lock to use git-config(1) locking, so it won't
113 # wait on the lock, unlike some of our internal flock()-based locks.
114 # This is to prevent direct git-config(1) usage from clobbering our
116 my $lockfile = "$pi_config.lock";
118 sysopen($lockfh, $lockfile, O_RDWR|O_CREAT|O_EXCL) or do {
119 warn "could not open config file: $lockfile: $!\n";
122 require PublicInbox::OnDestroy;
123 my $auto_unlink = PublicInbox::OnDestroy->new($$, sub { unlink $lockfile });
126 open(my $oh, '<', $pi_config) or die "unable to read $pi_config: $!\n";
129 defined $perm or die "(f)stat failed on $pi_config: $!\n";
130 chmod($perm & 07777, $fh) or
131 die "(f)chmod failed on future $pi_config: $!\n";
132 defined(my $old = do { local $/; <$oh> }) or die "read $pi_config: $!\n";
133 print $fh $old or die "failed to write: $!\n";
134 close $oh or die "failed to close $pi_config: $!\n";
136 # yes, this conflict checking is racy if multiple instances of this
137 # script are run by the same $PI_DIR
138 my $cfg = PublicInbox::Config->new;
140 foreach my $addr (@address) {
141 my $found = $cfg->lookup($addr);
143 if ($found->{name} ne $name) {
145 "`$addr' already defined for ",
146 "`$found->{name}',\n",
147 "does not match intend `$name'\n";
150 $seen{lc($addr)} = 1;
155 exit(1) if $conflict;
157 my $ibx = $cfg->lookup_name($name);
158 $indexlevel //= $ibx->{indexlevel} if $ibx;
160 my $pi_config_tmp = $fh->filename;
161 close($fh) or die "failed to close $pi_config_tmp: $!\n";
163 my $pfx = "publicinbox.$name";
164 my @x = (qw/git config/, "--file=$pi_config_tmp");
166 $inboxdir = PublicInbox::Config::rel2abs_collapsed($inboxdir);
167 die "`\\n' not allowed in `$inboxdir'\n" if index($inboxdir, "\n") >= 0;
169 if (-f "$inboxdir/inbox.lock") {
170 if (!defined $version) {
172 } elsif ($version != 2) {
173 die "$inboxdir is a -V2 inbox, -V$version specified\n"
175 } elsif (-d "$inboxdir/objects") {
176 if (!defined $version) {
178 } elsif ($version != 1) {
179 die "$inboxdir is a -V1 inbox, -V$version specified\n"
183 $version = 1 unless defined $version;
185 if ($version == 1 && defined $skip_epoch) {
186 die "--skip-epoch is only supported for -V2 inboxes\n";
189 my $ibx = PublicInbox::Inbox->new({
190 inboxdir => $inboxdir,
193 -primary_address => $address[0],
194 indexlevel => $indexlevel,
199 die "--jobs is only supported for -V2 inboxes\n" if $version == 1;
200 die "--jobs=$jobs must be >= 1\n" if $jobs <= 0;
201 $creat_opt->{nproc} = $jobs;
204 require PublicInbox::InboxWritable;
205 $ibx = PublicInbox::InboxWritable->new($ibx, $creat_opt);
207 $ibx->{indexlevel} //= 'full'; # ensure init_inbox writes xdb
208 $ibx->{indexlevel} eq 'basic' and
209 die "--skip-docdata ignored with --indexlevel=basic\n";
210 $ibx->{-skip_docdata} = $skip_docdata;
212 $ibx->init_inbox(0, $skip_epoch, $skip_artnum);
214 # needed for git prior to v2.1.0
215 umask(0077) if defined $perm;
217 require PublicInbox::Spawn;
218 PublicInbox::Spawn->import(qw(run_die));
220 foreach my $addr (@address) {
221 next if $seen{lc($addr)};
222 run_die([@x, "--add", "$pfx.address", $addr]);
224 run_die([@x, "$pfx.url", $http_url]);
225 run_die([@x, "$pfx.inboxdir", $inboxdir]);
227 if (defined($indexlevel)) {
228 run_die([@x, "$pfx.indexlevel", $indexlevel]);
230 run_die([@x, "$pfx.newsgroup", $ng]) if $ng ne '';
232 for my $kv (@c_extra) {
233 my ($k, $v) = split(/=/, $kv, 2);
234 # --fixed-value for idempotent invocations
235 run_die([@x, qw(--replace-all --fixed-value), "$pfx.$k", $v, $v]);
238 # needed for git prior to v2.1.0
240 chmod($perm & 07777, $pi_config_tmp) or
241 die "(f)chmod failed on future $pi_config: $!\n";
244 rename $pi_config_tmp, $pi_config or
245 die "failed to rename `$pi_config_tmp' to `$pi_config': $!\n";
246 undef $auto_unlink; # trigger ->DESTROY