1 # Copyright (C) 2016-2019 all contributors <meta@public-inbox.org>
2 # License: AGPL-3.0+ <https://www.gnu.org/licenses/agpl-3.0.txt>
4 # Represents a public-inbox (which may have multiple mailing addresses)
5 package PublicInbox::Inbox;
9 use PublicInbox::MID qw(mid2path);
10 use PublicInbox::MIME;
12 # Long-running "git-cat-file --batch" processes won't notice
13 # unlinked packs, so we need to restart those processes occasionally.
14 # Xapian and SQLite file handles are mostly stable, but sometimes an
15 # admin will attempt to replace them atomically after compact/vacuum
16 # and we need to be prepared for that.
18 my $cleanup_avail = -1; # 0, or 1
20 my $CLEANUP = {}; # string(inbox) -> inbox
22 $cleanup_timer = undef;
24 for my $ibx (values %$CLEANUP) {
26 if ($have_devel_peek) {
27 foreach my $f (qw(mm search over)) {
28 # we bump refcnt by assigning tmp, here:
29 my $tmp = $ibx->{$f} or next;
30 next if Devel::Peek::SvREFCNT($tmp) > 2;
32 # refcnt is zero when tmp is out-of-scope
35 if (my $git = $ibx->{git}) {
36 $again = $git->cleanup;
38 if (my $gits = $ibx->{-repo_objs}) {
39 foreach my $git (@$gits) {
40 $again = 1 if $git->cleanup;
43 if ($have_devel_peek) {
44 $again ||= !!($ibx->{over} || $ibx->{mm} ||
47 $next->{"$ibx"} = $ibx if $again;
52 sub cleanup_possible () {
53 # no need to require EvCleanup, here, if it were enabled another
54 # module would've require'd it, already
55 eval { PublicInbox::EvCleanup::enabled() } or return 0;
58 require Devel::Peek; # needs separate package in Fedora
64 sub _cleanup_later ($) {
66 $cleanup_avail = cleanup_possible() if $cleanup_avail < 0;
67 return if $cleanup_avail != 1;
68 $cleanup_timer ||= PublicInbox::EvCleanup::later(*cleanup_task);
69 $CLEANUP->{"$self"} = $self;
73 my ($opts, $field, $default) = @_;
74 my $val = $opts->{$field};
76 $val = $val->[-1] if ref($val) eq 'ARRAY';
77 $val = undef if $val !~ /\A[0-9]+\z/;
79 $opts->{$field} = $val || $default;
82 sub _set_limiter ($$$) {
83 my ($self, $pi_config, $pfx) = @_;
84 my $lkey = "-${pfx}_limiter";
85 $self->{$lkey} ||= eval {
86 # full key is: publicinbox.$NAME.httpbackendmax
87 my $mkey = $pfx.'max';
88 my $val = $self->{$mkey} or return;
90 if ($val =~ /\A[0-9]+\z/) {
91 require PublicInbox::Qspawn;
92 $lim = PublicInbox::Qspawn::Limiter->new($val);
93 } elsif ($val =~ /\A[a-z][a-z0-9]*\z/) {
94 $lim = $pi_config->limiter($val);
95 warn "$mkey limiter=$val not found\n" if !$lim;
97 warn "$mkey limiter=$val not understood\n";
104 my ($class, $opts) = @_;
105 my $v = $opts->{address} ||= 'public-inbox@example.com';
106 my $p = $opts->{-primary_address} = ref($v) eq 'ARRAY' ? $v->[0] : $v;
107 $opts->{domain} = ($p =~ /\@(\S+)\z/) ? $1 : 'localhost';
108 my $pi_config = delete $opts->{-pi_config};
109 _set_limiter($opts, $pi_config, 'httpbackend');
110 _set_uint($opts, 'feedmax', 25);
111 $opts->{nntpserver} ||= $pi_config->{'publicinbox.nntpserver'};
112 my $dir = $opts->{inboxdir};
113 if (defined $dir && -f "$dir/inbox.lock") {
114 $opts->{version} = 2;
117 # allow any combination of multi-line or comma-delimited hide entries
119 if (defined(my $h = $opts->{hide})) {
120 foreach my $v (@$h) {
121 $hide->{$_} = 1 foreach (split(/\s*,\s*/, $v));
123 $opts->{-hide} = $hide;
129 my ($self, $epoch) = @_;
130 ($self->{version} || 1) == 2 or return;
131 $self->{"$epoch.git"} ||= eval {
132 my $git_dir = "$self->{inboxdir}/git/$epoch.git";
133 my $g = PublicInbox::Git->new($git_dir);
134 $g->{-httpbackend_limiter} = $self->{-httpbackend_limiter};
135 # no cleanup needed, we never cat-file off this, only clone
142 $self->{git} ||= eval {
143 my $git_dir = $self->{inboxdir};
144 $git_dir .= '/all.git' if (($self->{version} || 1) == 2);
145 my $g = PublicInbox::Git->new($git_dir);
146 $g->{-httpbackend_limiter} = $self->{-httpbackend_limiter};
147 _cleanup_later($self);
154 my $v = $self->{version};
155 return unless defined($v) && $v == 2;
156 my $cur = $self->{-max_git_epoch};
157 my $changed = git($self)->alternates_changed;
158 if (!defined($cur) || $changed) {
159 $self->git->cleanup if $changed;
160 my $gits = "$self->{inboxdir}/git";
161 if (opendir my $dh, $gits) {
163 while (defined(my $git_dir = readdir($dh))) {
164 $git_dir =~ m!\A([0-9]+)\.git\z! or next;
165 $max = $1 if $1 > $max;
167 $cur = $self->{-max_git_epoch} = $max if $max >= 0;
169 warn "opendir $gits failed: $!\n";
177 $self->{mm} ||= eval {
178 require PublicInbox::Msgmap;
179 _cleanup_later($self);
180 my $dir = $self->{inboxdir};
181 if (($self->{version} || 1) >= 2) {
182 PublicInbox::Msgmap->new_file("$dir/msgmap.sqlite3");
184 PublicInbox::Msgmap->new($dir);
190 my ($self, $over_only) = @_;
191 my $srch = $self->{search} ||= eval {
192 _cleanup_later($self);
193 require PublicInbox::Search;
194 PublicInbox::Search->new($self);
196 ($over_only || eval { $srch->xdb }) ? $srch : undef;
201 my $srch = search($self, 1) or return;
202 $self->{over} ||= eval {
203 my $over = $srch->{over_ro};
204 $over->dbh_new; # may fail
212 if (open(my $fh, '<', $path)) {
221 my $desc = $self->{description};
222 return $desc if defined $desc;
223 $desc = try_cat("$self->{inboxdir}/description");
226 $desc =~ s/\s+/ /smg;
227 $desc = '($INBOX_DIR/description missing)' if $desc eq '';
228 $self->{description} = $desc;
233 my $url = $self->{cloneurl};
235 $url = try_cat("$self->{inboxdir}/cloneurl");
236 my @url = split(/\s+/s, $url);
239 $self->{cloneurl} = \@url;
243 my ($self, $env) = @_;
245 if ($env && ($scheme = $env->{'psgi.url_scheme'})) { # PSGI env
246 my $host_port = $env->{HTTP_HOST} ||
247 "$env->{SERVER_NAME}:$env->{SERVER_PORT}";
248 my $url = "$scheme://$host_port". ($env->{SCRIPT_NAME} || '/');
249 # for mount in Plack::Builder
250 $url .= '/' if $url !~ m!/\z!;
251 $url .= $self->{name} . '/';
253 # either called from a non-PSGI environment (e.g. NNTP/POP3)
254 $self->{-base_url} ||= do {
255 my $url = $self->{url} or return undef;
256 # expand protocol-relative URLs to HTTPS if we're
257 # not inside a web server
258 $url = "https:$url" if $url =~ m!\A//!;
259 $url .= '/' if $url !~ m!/\z!;
267 $self->{-nntp_url} ||= do {
268 # no checking for nntp_usable here, we can point entirely
269 # to non-local servers or users run by a different user
270 my $ns = $self->{nntpserver};
271 my $group = $self->{newsgroup};
274 $ns = [ $ns ] if ref($ns) ne 'ARRAY';
276 my $u = m!\Anntps?://! ? $_ : "nntp://$_";
277 $u .= '/' if $u !~ m!/\z!;
282 my $mirrors = $self->{nntpmirror};
285 foreach (@$mirrors) {
286 my $u = m!\Anntps?://! ? $_ : "nntp://$_";
287 if ($u =~ m!\Anntps?://[^/]+/?\z!) {
289 $u .= '/' if $u !~ m!/\z!;
293 "publicinbox.$self->{name}.nntpmirror=$_ missing newsgroup name\n";
296 # else: allow full URLs like:
297 # nntp://news.example.com/alt.example
300 my %seen = map { $_ => 1 } @urls;
313 my $ret = mm($self) && over($self);
314 $self->{mm} = $self->{over} = $self->{search} = undef;
318 sub msg_by_path ($$;$) {
319 my ($self, $path, $ref) = @_;
320 # TODO: allow other refs:
321 my $str = git($self)->cat_file('HEAD:'.$path, $ref);
322 $$str =~ s/\A[\r\n]*From [^\r\n]*\r?\n//s if $str;
326 sub msg_by_smsg ($$;$) {
327 my ($self, $smsg, $ref) = @_;
329 # ghosts may have undef smsg (from SearchThread.node) or
331 return unless defined $smsg;
332 defined(my $blob = $smsg->{blob}) or return;
334 my $str = git($self)->cat_file($blob, $ref);
335 $$str =~ s/\A[\r\n]*From [^\r\n]*\r?\n//s if $str;
340 my ($self, $smsg, $ref) = @_;
341 if (my $s = msg_by_smsg($self, $smsg, $ref)) {
342 $smsg->{mime} = PublicInbox::MIME->new($s);
348 my ($self, $mid) = @_;
349 my $mm = mm($self) or return;
353 sub smsg_by_mid ($$) {
354 my ($self, $mid) = @_;
355 my $over = over($self) or return;
356 # favor the Message-ID we used for the NNTP article number:
357 defined(my $num = mid2num($self, $mid)) or return;
358 my $smsg = $over->get_art($num) or return;
359 PublicInbox::SearchMsg::psgi_cull($smsg);
362 sub msg_by_mid ($$;$) {
363 my ($self, $mid, $ref) = @_;
366 return msg_by_path($self, mid2path($mid), $ref);
368 my $smsg = smsg_by_mid($self, $mid);
369 $smsg ? msg_by_smsg($self, $smsg, $ref) : undef;
373 my ($self, $opts, $after, $before) = @_;
374 over($self)->recent($opts, $after, $before);
379 if (my $over = over($self)) {
380 my $msgs = $over->recent({limit => 1});
381 if (my $smsg = $msgs->[0]) {
386 git($self)->modified; # v1