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 $pi_config = PublicInbox::Config->default_file;
96 my ($dir) = ($pi_config =~ m!(.*?/)[^/]+\z!);
98 File::Path::mkpath($dir); # will croak on fatal errors
100 # first, we grab a flock to prevent simultaneous public-inbox-init
101 # processes from trampling over each other, or exiting with 255 on
102 # O_EXCL failure below. This gets unlocked automatically on exit:
103 require PublicInbox::Lock;
104 my $lock_obj = { lock_path => "$pi_config.flock" };
105 PublicInbox::Lock::lock_acquire($lock_obj);
107 # git-config will operate on this (and rename on success):
109 my $fh = File::Temp->new(TEMPLATE => 'pi-init-XXXX', DIR => $dir);
111 # Now, we grab another lock to use git-config(1) locking, so it won't
112 # wait on the lock, unlike some of our internal flock()-based locks.
113 # This is to prevent direct git-config(1) usage from clobbering our
115 my $lockfile = "$pi_config.lock";
117 sysopen($lockfh, $lockfile, O_RDWR|O_CREAT|O_EXCL) or do {
118 warn "could not open config file: $lockfile: $!\n";
121 require PublicInbox::OnDestroy;
122 my $auto_unlink = PublicInbox::OnDestroy->new($$, sub { unlink $lockfile });
125 open(my $oh, '<', $pi_config) or die "unable to read $pi_config: $!\n";
128 defined $perm or die "(f)stat failed on $pi_config: $!\n";
129 chmod($perm & 07777, $fh) or
130 die "(f)chmod failed on future $pi_config: $!\n";
131 defined(my $old = do { local $/; <$oh> }) or die "read $pi_config: $!\n";
132 print $fh $old or die "failed to write: $!\n";
133 close $oh or die "failed to close $pi_config: $!\n";
135 # yes, this conflict checking is racy if multiple instances of this
136 # script are run by the same $PI_DIR
137 my $cfg = PublicInbox::Config->new;
139 foreach my $addr (@address) {
140 my $found = $cfg->lookup($addr);
142 if ($found->{name} ne $name) {
144 "`$addr' already defined for ",
145 "`$found->{name}',\n",
146 "does not match intend `$name'\n";
149 $seen{lc($addr)} = 1;
154 exit(1) if $conflict;
156 my $ibx = $cfg->lookup_name($name);
157 $indexlevel //= $ibx->{indexlevel} if $ibx;
159 my $pi_config_tmp = $fh->filename;
160 close($fh) or die "failed to close $pi_config_tmp: $!\n";
162 my $pfx = "publicinbox.$name";
163 my @x = (qw/git config/, "--file=$pi_config_tmp");
165 $inboxdir = PublicInbox::Config::rel2abs_collapsed($inboxdir);
166 die "`\\n' not allowed in `$inboxdir'\n" if index($inboxdir, "\n") >= 0;
168 if (-f "$inboxdir/inbox.lock") {
169 if (!defined $version) {
171 } elsif ($version != 2) {
172 die "$inboxdir is a -V2 inbox, -V$version specified\n"
174 } elsif (-d "$inboxdir/objects") {
175 if (!defined $version) {
177 } elsif ($version != 1) {
178 die "$inboxdir is a -V1 inbox, -V$version specified\n"
182 $version = 1 unless defined $version;
184 if ($version == 1 && defined $skip_epoch) {
185 die "--skip-epoch is only supported for -V2 inboxes\n";
188 my $ibx = PublicInbox::Inbox->new({
189 inboxdir => $inboxdir,
192 -primary_address => $address[0],
193 indexlevel => $indexlevel,
198 die "--jobs is only supported for -V2 inboxes\n" if $version == 1;
199 die "--jobs=$jobs must be >= 1\n" if $jobs <= 0;
200 $creat_opt->{nproc} = $jobs;
203 require PublicInbox::InboxWritable;
204 $ibx = PublicInbox::InboxWritable->new($ibx, $creat_opt);
206 $ibx->{indexlevel} //= 'full'; # ensure init_inbox writes xdb
207 $ibx->{indexlevel} eq 'basic' and
208 die "--skip-docdata ignored with --indexlevel=basic\n";
209 $ibx->{-skip_docdata} = $skip_docdata;
211 $ibx->init_inbox(0, $skip_epoch, $skip_artnum);
213 # needed for git prior to v2.1.0
214 umask(0077) if defined $perm;
216 require PublicInbox::Spawn;
217 PublicInbox::Spawn->import(qw(run_die));
219 foreach my $addr (@address) {
220 next if $seen{lc($addr)};
221 run_die([@x, "--add", "$pfx.address", $addr]);
223 run_die([@x, "$pfx.url", $http_url]);
224 run_die([@x, "$pfx.inboxdir", $inboxdir]);
226 if (defined($indexlevel)) {
227 run_die([@x, "$pfx.indexlevel", $indexlevel]);
229 run_die([@x, "$pfx.newsgroup", $ng]) if $ng ne '';
231 for my $kv (@c_extra) {
232 my ($k, $v) = split(/=/, $kv, 2);
233 # git 2.30+ has --fixed-value for idempotent invocations,
234 # but that's too new to depend on in 2021. Perl quotemeta
235 # seems compatible enough for POSIX ERE which git uses
236 my $re = '^'.quotemeta($v).'$';
237 run_die([@x, qw(--replace-all), "$pfx.$k", $v, $re]);
240 # needed for git prior to v2.1.0
242 chmod($perm & 07777, $pi_config_tmp) or
243 die "(f)chmod failed on future $pi_config: $!\n";
246 rename $pi_config_tmp, $pi_config or
247 die "failed to rename `$pi_config_tmp' to `$pi_config': $!\n";
248 undef $auto_unlink; # trigger ->DESTROY
250 my $f = "$inboxdir/description";
251 if (sysopen $fh, $f, O_CREAT|O_EXCL|O_WRONLY) {
252 print $fh "public inbox for $address[0]\n" or die "print($f): $!";
253 close $fh or die "close($f): $!";