1 # Copyright (C) 2016-2018 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 my $expire = time - 60;
36 if (my $git = $ibx->{git}) {
37 $again = $git->cleanup($expire);
39 if (my $gits = $ibx->{-repo_objs}) {
40 foreach my $git (@$gits) {
41 $again = 1 if $git->cleanup($expire);
44 if ($have_devel_peek) {
45 $again ||= !!($ibx->{over} || $ibx->{mm} ||
48 $next->{"$ibx"} = $ibx if $again;
53 sub cleanup_possible () {
54 # no need to require EvCleanup, here, if it were enabled another
55 # module would've require'd it, already
56 eval { PublicInbox::EvCleanup::enabled() } or return 0;
59 require Devel::Peek; # needs separate package in Fedora
65 sub _cleanup_later ($) {
67 $cleanup_avail = cleanup_possible() if $cleanup_avail < 0;
68 return if $cleanup_avail != 1;
69 $cleanup_timer ||= PublicInbox::EvCleanup::later(*cleanup_task);
70 $CLEANUP->{"$self"} = $self;
74 my ($opts, $field, $default) = @_;
75 my $val = $opts->{$field};
77 $val = $val->[-1] if ref($val) eq 'ARRAY';
78 $val = undef if $val !~ /\A\d+\z/;
80 $opts->{$field} = $val || $default;
83 sub _set_limiter ($$$) {
84 my ($self, $pi_config, $pfx) = @_;
85 my $lkey = "-${pfx}_limiter";
86 $self->{$lkey} ||= eval {
87 # full key is: publicinbox.$NAME.httpbackendmax
88 my $mkey = $pfx.'max';
89 my $val = $self->{$mkey} or return;
91 if ($val =~ /\A\d+\z/) {
92 require PublicInbox::Qspawn;
93 $lim = PublicInbox::Qspawn::Limiter->new($val);
94 } elsif ($val =~ /\A[a-z][a-z0-9]*\z/) {
95 $lim = $pi_config->limiter($val);
96 warn "$mkey limiter=$val not found\n" if !$lim;
98 warn "$mkey limiter=$val not understood\n";
105 my ($class, $opts) = @_;
106 my $v = $opts->{address} ||= 'public-inbox@example.com';
107 my $p = $opts->{-primary_address} = ref($v) eq 'ARRAY' ? $v->[0] : $v;
108 $opts->{domain} = ($p =~ /\@(\S+)\z/) ? $1 : 'localhost';
109 my $pi_config = delete $opts->{-pi_config};
110 _set_limiter($opts, $pi_config, 'httpbackend');
111 _set_uint($opts, 'feedmax', 25);
112 $opts->{nntpserver} ||= $pi_config->{'publicinbox.nntpserver'};
113 my $dir = $opts->{mainrepo};
114 if (defined $dir && -f "$dir/inbox.lock") {
115 $opts->{version} = 2;
118 # allow any combination of multi-line or comma-delimited hide entries
120 if (defined(my $h = $opts->{hide})) {
121 foreach my $v (@$h) {
122 $hide->{$_} = 1 foreach (split(/\s*,\s*/, $v));
124 $opts->{-hide} = $hide;
130 my ($self, $part) = @_;
131 ($self->{version} || 1) == 2 or return;
132 $self->{"$part.git"} ||= eval {
133 my $git_dir = "$self->{mainrepo}/git/$part.git";
134 my $g = PublicInbox::Git->new($git_dir);
135 $g->{-httpbackend_limiter} = $self->{-httpbackend_limiter};
136 # no cleanup needed, we never cat-file off this, only clone
143 $self->{git} ||= eval {
144 my $git_dir = $self->{mainrepo};
145 $git_dir .= '/all.git' if (($self->{version} || 1) == 2);
146 my $g = PublicInbox::Git->new($git_dir);
147 $g->{-httpbackend_limiter} = $self->{-httpbackend_limiter};
148 _cleanup_later($self);
155 my $v = $self->{version};
156 return unless defined($v) && $v == 2;
157 my $part = $self->{-max_git_part};
158 my $changed = git($self)->alternates_changed;
159 if (!defined($part) || $changed) {
160 $self->git->cleanup if $changed;
161 my $gits = "$self->{mainrepo}/git";
162 if (opendir my $dh, $gits) {
164 while (defined(my $git_dir = readdir($dh))) {
165 $git_dir =~ m!\A(\d+)\.git\z! or next;
166 $max = $1 if $1 > $max;
168 $part = $self->{-max_git_part} = $max if $max >= 0;
170 warn "opendir $gits failed: $!\n";
178 $self->{mm} ||= eval {
179 require PublicInbox::Msgmap;
180 _cleanup_later($self);
181 my $dir = $self->{mainrepo};
182 if (($self->{version} || 1) >= 2) {
183 PublicInbox::Msgmap->new_file("$dir/msgmap.sqlite3");
185 PublicInbox::Msgmap->new($dir);
191 my ($self, $over_only) = @_;
192 my $srch = $self->{search} ||= eval {
193 _cleanup_later($self);
194 require PublicInbox::Search;
195 PublicInbox::Search->new($self, $self->{altid});
197 ($over_only || eval { $srch->xdb }) ? $srch : undef;
202 my $srch = search($self, 1) or return;
203 $self->{over} ||= eval {
204 my $over = $srch->{over_ro};
205 $over->dbh_new; # may fail
213 if (open(my $fh, '<', $path)) {
222 my $desc = $self->{description};
223 return $desc if defined $desc;
224 $desc = try_cat("$self->{mainrepo}/description");
227 $desc =~ s/\s+/ /smg;
228 $desc = '($INBOX_DIR/description missing)' if $desc eq '';
229 $self->{description} = $desc;
234 my $url = $self->{cloneurl};
236 $url = try_cat("$self->{mainrepo}/cloneurl");
237 my @url = split(/\s+/s, $url);
240 $self->{cloneurl} = \@url;
244 my ($self, $env) = @_;
246 if ($env && ($scheme = $env->{'psgi.url_scheme'})) { # PSGI env
247 my $host_port = $env->{HTTP_HOST} ||
248 "$env->{SERVER_NAME}:$env->{SERVER_PORT}";
249 my $url = "$scheme://$host_port". ($env->{SCRIPT_NAME} || '/');
250 # for mount in Plack::Builder
251 $url .= '/' if $url !~ m!/\z!;
252 $url .= $self->{name} . '/';
254 # either called from a non-PSGI environment (e.g. NNTP/POP3)
255 $self->{-base_url} ||= do {
256 my $url = $self->{url} or return undef;
257 # expand protocol-relative URLs to HTTPS if we're
258 # not inside a web server
259 $url = "https:$url" if $url =~ m!\A//!;
260 $url .= '/' if $url !~ m!/\z!;
268 $self->{-nntp_url} ||= do {
269 # no checking for nntp_usable here, we can point entirely
270 # to non-local servers or users run by a different user
271 my $ns = $self->{nntpserver};
272 my $group = $self->{newsgroup};
275 $ns = [ $ns ] if ref($ns) ne 'ARRAY';
277 my $u = m!\Anntps?://! ? $_ : "nntp://$_";
278 $u .= '/' if $u !~ m!/\z!;
283 my $mirrors = $self->{nntpmirror};
286 foreach (@$mirrors) {
287 my $u = m!\Anntps?://! ? $_ : "nntp://$_";
288 if ($u =~ m!\Anntps?://[^/]+/?\z!) {
290 $u .= '/' if $u !~ m!/\z!;
294 "publicinbox.$self->{name}.nntpmirror=$_ missing newsgroup name\n";
297 # else: allow full URLs like:
298 # nntp://news.example.com/alt.example
301 my %seen = map { $_ => 1 } @urls;
314 my $ret = mm($self) && over($self);
315 $self->{mm} = $self->{over} = $self->{search} = undef;
319 sub msg_by_path ($$;$) {
320 my ($self, $path, $ref) = @_;
321 # TODO: allow other refs:
322 my $str = git($self)->cat_file('HEAD:'.$path, $ref);
323 $$str =~ s/\A[\r\n]*From [^\r\n]*\r?\n//s if $str;
327 sub msg_by_smsg ($$;$) {
328 my ($self, $smsg, $ref) = @_;
330 # ghosts may have undef smsg (from SearchThread.node) or
332 return unless defined $smsg;
333 defined(my $blob = $smsg->{blob}) or return;
335 my $str = git($self)->cat_file($blob, $ref);
336 $$str =~ s/\A[\r\n]*From [^\r\n]*\r?\n//s if $str;
341 my ($self, $smsg, $ref) = @_;
342 if (my $s = msg_by_smsg($self, $smsg, $ref)) {
343 $smsg->{mime} = PublicInbox::MIME->new($s);
349 my ($self, $mid) = @_;
350 my $mm = mm($self) or return;
354 sub smsg_by_mid ($$) {
355 my ($self, $mid) = @_;
356 my $over = over($self) or return;
357 # favor the Message-ID we used for the NNTP article number:
358 defined(my $num = mid2num($self, $mid)) or return;
359 my $smsg = $over->get_art($num) or return;
360 PublicInbox::SearchMsg::psgi_cull($smsg);
363 sub msg_by_mid ($$;$) {
364 my ($self, $mid, $ref) = @_;
367 return msg_by_path($self, mid2path($mid), $ref);
369 my $smsg = smsg_by_mid($self, $mid);
370 $smsg ? msg_by_smsg($self, $smsg, $ref) : undef;
374 my ($self, $opts, $after, $before) = @_;
375 over($self)->recent($opts, $after, $before);
380 if (my $over = over($self)) {
381 my $msgs = $over->recent({limit => 1});
382 if (my $smsg = $msgs->[0]) {
387 git($self)->modified; # v1