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);
39 my (@c_extra, @chdir);
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,
56 GetOptions(%opts) or $usage_cb->();
57 if ($show_help) { print $help; exit 0 };
58 my $name = shift @ARGV or $usage_cb->();
59 my $inboxdir = shift @ARGV or $usage_cb->();
60 my $http_url = shift @ARGV or $usage_cb->();
61 my (@address) = @ARGV;
62 @address or $usage_cb->();
63 +PublicInbox::Admin::do_chdir(\@chdir);
66 my ($k, $v) = split(/=/, $_, 2);
67 defined($v) or die "Usage: -c KEY=VALUE\n";
68 $k =~ /\A[a-z]+\z/i or die "$k contains invalid characters\n";
70 if ($k eq 'newsgroup') {
71 die "newsgroup already set ($ng)\n" if $ng ne '';
74 } elsif ($k eq 'address') {
75 push @address, $v; # for conflict checking
77 } elsif ($k =~ /\A(?:inboxdir|mainrepo)\z/) {
78 die "$k not allowed via -c $_\n"
79 } elsif ($k eq 'indexlevel') {
80 defined($indexlevel) and
81 die "indexlevel already set ($indexlevel)\n";
89 PublicInbox::Admin::indexlevel_ok_or_die($indexlevel) if defined $indexlevel;
91 $ng =~ m![^A-Za-z0-9/_\.\-\~\@\+\=:]! and
92 die "--newsgroup `$ng' is not valid\n";
93 ($ng =~ m!\A\.! || $ng =~ m!\.\z!) and
94 die "--newsgroup `$ng' must not start or end with `.'\n";
96 require PublicInbox::Config;
97 my $pi_config = PublicInbox::Config->default_file;
98 my ($dir) = ($pi_config =~ m!(.*?/)[^/]+\z!);
100 File::Path::mkpath($dir); # will croak on fatal errors
102 # first, we grab a flock to prevent simultaneous public-inbox-init
103 # processes from trampling over each other, or exiting with 255 on
104 # O_EXCL failure below. This gets unlocked automatically on exit:
105 require PublicInbox::Lock;
106 my $lock_obj = { lock_path => "$pi_config.flock" };
107 PublicInbox::Lock::lock_acquire($lock_obj);
109 # git-config will operate on this (and rename on success):
111 my $fh = File::Temp->new(TEMPLATE => 'pi-init-XXXX', DIR => $dir);
113 # Now, we grab another lock to use git-config(1) locking, so it won't
114 # wait on the lock, unlike some of our internal flock()-based locks.
115 # This is to prevent direct git-config(1) usage from clobbering our
117 my $lockfile = "$pi_config.lock";
119 sysopen($lockfh, $lockfile, O_RDWR|O_CREAT|O_EXCL) or do {
120 warn "could not open config file: $lockfile: $!\n";
123 require PublicInbox::OnDestroy;
124 my $auto_unlink = PublicInbox::OnDestroy->new($$, sub { unlink $lockfile });
127 open(my $oh, '<', $pi_config) or die "unable to read $pi_config: $!\n";
130 defined $perm or die "(f)stat failed on $pi_config: $!\n";
131 chmod($perm & 07777, $fh) or
132 die "(f)chmod failed on future $pi_config: $!\n";
133 defined(my $old = do { local $/; <$oh> }) or die "read $pi_config: $!\n";
134 print $fh $old or die "failed to write: $!\n";
135 close $oh or die "failed to close $pi_config: $!\n";
137 # yes, this conflict checking is racy if multiple instances of this
138 # script are run by the same $PI_DIR
139 my $cfg = PublicInbox::Config->new;
141 foreach my $addr (@address) {
142 my $found = $cfg->lookup($addr);
144 if ($found->{name} ne $name) {
146 "`$addr' already defined for ",
147 "`$found->{name}',\n",
148 "does not match intend `$name'\n";
151 $seen{lc($addr)} = 1;
156 exit(1) if $conflict;
158 my $ibx = $cfg->lookup_name($name);
159 $indexlevel //= $ibx->{indexlevel} if $ibx;
161 my $pi_config_tmp = $fh->filename;
162 close($fh) or die "failed to close $pi_config_tmp: $!\n";
164 my $pfx = "publicinbox.$name";
165 my @x = (qw/git config/, "--file=$pi_config_tmp");
167 $inboxdir = PublicInbox::Config::rel2abs_collapsed($inboxdir);
168 die "`\\n' not allowed in `$inboxdir'\n" if index($inboxdir, "\n") >= 0;
170 if (-f "$inboxdir/inbox.lock") {
171 if (!defined $version) {
173 } elsif ($version != 2) {
174 die "$inboxdir is a -V2 inbox, -V$version specified\n"
176 } elsif (-d "$inboxdir/objects") {
177 if (!defined $version) {
179 } elsif ($version != 1) {
180 die "$inboxdir is a -V1 inbox, -V$version specified\n"
184 $version = 1 unless defined $version;
186 if ($version == 1 && defined $skip_epoch) {
187 die "--skip-epoch is only supported for -V2 inboxes\n";
190 my $ibx = PublicInbox::Inbox->new({
191 inboxdir => $inboxdir,
194 -primary_address => $address[0],
195 indexlevel => $indexlevel,
200 die "--jobs is only supported for -V2 inboxes\n" if $version == 1;
201 die "--jobs=$jobs must be >= 1\n" if $jobs <= 0;
202 $creat_opt->{nproc} = $jobs;
205 require PublicInbox::InboxWritable;
206 $ibx = PublicInbox::InboxWritable->new($ibx, $creat_opt);
208 $ibx->{indexlevel} //= 'full'; # ensure init_inbox writes xdb
209 $ibx->{indexlevel} eq 'basic' and
210 die "--skip-docdata ignored with --indexlevel=basic\n";
211 $ibx->{-skip_docdata} = $skip_docdata;
213 $ibx->init_inbox(0, $skip_epoch, $skip_artnum);
215 # needed for git prior to v2.1.0
216 umask(0077) if defined $perm;
218 require PublicInbox::Spawn;
219 PublicInbox::Spawn->import(qw(run_die));
221 foreach my $addr (@address) {
222 next if $seen{lc($addr)};
223 run_die([@x, "--add", "$pfx.address", $addr]);
225 run_die([@x, "$pfx.url", $http_url]);
226 run_die([@x, "$pfx.inboxdir", $inboxdir]);
228 if (defined($indexlevel)) {
229 run_die([@x, "$pfx.indexlevel", $indexlevel]);
231 run_die([@x, "$pfx.newsgroup", $ng]) if $ng ne '';
233 for my $kv (@c_extra) {
234 my ($k, $v) = split(/=/, $kv, 2);
235 # git 2.30+ has --fixed-value for idempotent invocations,
236 # but that's too new to depend on in 2021. Perl quotemeta
237 # seems compatible enough for POSIX ERE which git uses
238 my $re = '^'.quotemeta($v).'$';
239 run_die([@x, qw(--replace-all), "$pfx.$k", $v, $re]);
242 # needed for git prior to v2.1.0
244 chmod($perm & 07777, $pi_config_tmp) or
245 die "(f)chmod failed on future $pi_config: $!\n";
248 rename $pi_config_tmp, $pi_config or
249 die "failed to rename `$pi_config_tmp' to `$pi_config': $!\n";
250 undef $auto_unlink; # trigger ->DESTROY
252 my $f = "$inboxdir/description";
253 if (sysopen $fh, $f, O_CREAT|O_EXCL|O_WRONLY) {
254 print $fh "public inbox for $address[0]\n" or die "print($f): $!";
255 close $fh or die "close($f): $!";