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)) {
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->{mm} || $ibx->{search});
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\d+\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\d+\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->{mainrepo};
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, $part) = @_;
130 ($self->{version} || 1) == 2 or return;
131 $self->{"$part.git"} ||= eval {
132 my $git_dir = "$self->{mainrepo}/git/$part.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->{mainrepo};
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 $part = $self->{-max_git_part};
157 my $changed = git($self)->alternates_changed;
158 if (!defined($part) || $changed) {
159 $self->git->cleanup if $changed;
160 my $gits = "$self->{mainrepo}/git";
161 if (opendir my $dh, $gits) {
163 while (defined(my $git_dir = readdir($dh))) {
164 $git_dir =~ m!\A(\d+)\.git\z! or next;
165 $max = $1 if $1 > $max;
167 $part = $self->{-max_git_part} = $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->{mainrepo};
181 if (($self->{version} || 1) >= 2) {
182 PublicInbox::Msgmap->new_file("$dir/msgmap.sqlite3");
184 PublicInbox::Msgmap->new($dir);
191 $self->{search} ||= eval {
192 _cleanup_later($self);
193 PublicInbox::Search->new($self, $self->{altid});
200 if (open(my $fh, '<', $path)) {
209 my $desc = $self->{description};
210 return $desc if defined $desc;
211 $desc = try_cat("$self->{mainrepo}/description");
214 $desc =~ s/\s+/ /smg;
215 $desc = '($REPO_DIR/description missing)' if $desc eq '';
216 $self->{description} = $desc;
221 my $url = $self->{cloneurl};
223 $url = try_cat("$self->{mainrepo}/cloneurl");
224 my @url = split(/\s+/s, $url);
227 $self->{cloneurl} = \@url;
231 my ($self, $env) = @_;
233 if ($env && ($scheme = $env->{'psgi.url_scheme'})) { # PSGI env
234 my $host_port = $env->{HTTP_HOST} ||
235 "$env->{SERVER_NAME}:$env->{SERVER_PORT}";
236 my $url = "$scheme://$host_port". ($env->{SCRIPT_NAME} || '/');
237 # for mount in Plack::Builder
238 $url .= '/' if $url !~ m!/\z!;
239 $url .= $self->{name} . '/';
241 # either called from a non-PSGI environment (e.g. NNTP/POP3)
242 $self->{-base_url} ||= do {
243 my $url = $self->{url} or return undef;
244 # expand protocol-relative URLs to HTTPS if we're
245 # not inside a web server
246 $url = "https:$url" if $url =~ m!\A//!;
247 $url .= '/' if $url !~ m!/\z!;
255 $self->{-nntp_url} ||= do {
256 # no checking for nntp_usable here, we can point entirely
257 # to non-local servers or users run by a different user
258 my $ns = $self->{nntpserver};
259 my $group = $self->{newsgroup};
262 $ns = [ $ns ] if ref($ns) ne 'ARRAY';
264 my $u = m!\Anntps?://! ? $_ : "nntp://$_";
265 $u .= '/' if $u !~ m!/\z!;
270 my $mirrors = $self->{nntpmirror};
273 foreach (@$mirrors) {
274 my $u = m!\Anntps?://! ? $_ : "nntp://$_";
275 if ($u =~ m!\Anntps?://[^/]+/?\z!) {
277 $u .= '/' if $u !~ m!/\z!;
281 "publicinbox.$self->{name}.nntpmirror=$_ missing newsgroup name\n";
284 # else: allow full URLs like:
285 # nntp://news.example.com/alt.example
288 my %seen = map { $_ => 1 } @urls;
301 my $ret = $self->mm && $self->search;
302 $self->{mm} = $self->{search} = undef;
306 sub msg_by_path ($$;$) {
307 my ($self, $path, $ref) = @_;
308 # TODO: allow other refs:
309 my $str = git($self)->cat_file('HEAD:'.$path, $ref);
310 $$str =~ s/\A[\r\n]*From [^\r\n]*\r?\n//s if $str;
314 sub msg_by_smsg ($$;$) {
315 my ($self, $smsg, $ref) = @_;
317 # ghosts may have undef smsg (from SearchThread.node) or
319 return unless defined $smsg;
320 defined(my $blob = $smsg->{blob}) or return;
322 my $str = git($self)->cat_file($blob, $ref);
323 $$str =~ s/\A[\r\n]*From [^\r\n]*\r?\n//s if $str;
328 my ($self, $smsg, $ref) = @_;
329 if (my $s = msg_by_smsg($self, $smsg, $ref)) {
330 $smsg->{mime} = PublicInbox::MIME->new($s);
336 my ($self, $mid) = @_;
337 my $mm = mm($self) or return;
341 sub smsg_by_mid ($$) {
342 my ($self, $mid) = @_;
343 my $srch = search($self) or return;
344 # favor the Message-ID we used for the NNTP article number:
345 defined(my $num = mid2num($self, $mid)) or return;
346 my $smsg = $srch->lookup_article($num) or return;
347 PublicInbox::SearchMsg::psgi_cull($smsg);
350 sub msg_by_mid ($$;$) {
351 my ($self, $mid, $ref) = @_;
352 my $srch = search($self) or
353 return msg_by_path($self, mid2path($mid), $ref);
354 my $smsg = smsg_by_mid($self, $mid);
355 $smsg ? msg_by_smsg($self, $smsg, $ref) : undef;
359 my ($self, $opts, $after, $before) = @_;
360 search($self)->{over_ro}->recent($opts, $after, $before);
365 if (my $srch = search($self)) {
366 my $msgs = $srch->{over_ro}->recent({limit => 1});
367 if (my $smsg = $msgs->[0]) {
372 git($self)->modified; # v1