-# Copyright (C) 2016-2020 all contributors <meta@public-inbox.org>
+# Copyright (C) 2016-2021 all contributors <meta@public-inbox.org>
# License: AGPL-3.0+ <https://www.gnu.org/licenses/agpl-3.0.txt>
#
# Represents a public-inbox (which may have multiple mailing addresses)
package PublicInbox::Inbox;
use strict;
-use warnings;
use PublicInbox::Git;
use PublicInbox::MID qw(mid2path);
use PublicInbox::Eml;
+use List::Util qw(max);
+use Carp qw(croak);
# Long-running "git-cat-file --batch" processes won't notice
# unlinked packs, so we need to restart those processes occasionally.
# admin will attempt to replace them atomically after compact/vacuum
# and we need to be prepared for that.
my $cleanup_timer;
-my $cleanup_avail = -1; # 0, or 1
-my $have_devel_peek;
my $CLEANUP = {}; # string(inbox) -> inbox
sub git_cleanup ($) {
my ($self) = @_;
- my $git = $self->{git} or return;
- $git->cleanup;
+ my $git = $self->{git} // return undef;
+ # normal inboxes have low startup cost and there may be many, so
+ # keep process+pipe counts in check. ExtSearch may have high startup
+ # cost (e.g. ->ALL) and but likely one per-daemon, so cleanup only
+ # if there's unlinked files
+ my $live = $self->isa(__PACKAGE__) ? $git->cleanup
+ : $git->cleanup_if_unlinked;
+ delete($self->{git}) unless $live;
+ $live;
}
+# returns true if further checking is required
+sub cleanup_shards { $_[0]->{search} ? $_[0]->{search}->cleanup_shards : undef }
+
sub cleanup_task () {
$cleanup_timer = undef;
my $next = {};
for my $ibx (values %$CLEANUP) {
- my $again;
- if ($have_devel_peek) {
- foreach my $f (qw(search)) {
- # we bump refcnt by assigning tmp, here:
- my $tmp = $ibx->{$f} or next;
- next if Devel::Peek::SvREFCNT($tmp) > 2;
- delete $ibx->{$f};
- # refcnt is zero when tmp is out-of-scope
- }
- }
- git_cleanup($ibx);
- if (my $gits = $ibx->{-repo_objs}) {
- foreach my $git (@$gits) {
- $again = 1 if $git->cleanup;
- }
+ my $again = git_cleanup($ibx);
+ $ibx->cleanup_shards and $again = 1;
+ for my $git (@{$ibx->{-repo_objs}}) {
+ $again = 1 if $git->cleanup;
}
check_inodes($ibx);
- if ($have_devel_peek) {
- $again ||= !!$ibx->{search};
- }
$next->{"$ibx"} = $ibx if $again;
}
$CLEANUP = $next;
+ $cleanup_timer //= PublicInbox::DS::later(\&cleanup_task);
}
-sub cleanup_possible () {
+sub _cleanup_later ($) {
# no need to require DS, here, if it were enabled another
# module would've require'd it, already
- eval { PublicInbox::DS::in_loop() } or return 0;
-
- eval {
- require Devel::Peek; # needs separate package in Fedora
- $have_devel_peek = 1;
- };
- 1;
-}
-
-sub _cleanup_later ($) {
- my ($self) = @_;
- $cleanup_avail = cleanup_possible() if $cleanup_avail < 0;
- return if $cleanup_avail != 1;
- $cleanup_timer //= PublicInbox::DS::later(\&cleanup_task);
- $CLEANUP->{"$self"} = $self;
-}
-
-sub _set_uint ($$$) {
- my ($opts, $field, $default) = @_;
- my $val = $opts->{$field};
- if (defined $val) {
- $val = $val->[-1] if ref($val) eq 'ARRAY';
- $val = undef if $val !~ /\A[0-9]+\z/;
+ if (eval { PublicInbox::DS::in_loop() }) {
+ $cleanup_timer //= PublicInbox::DS::later(\&cleanup_task);
+ $CLEANUP->{"$_[0]"} = $_[0]; # $self
}
- $opts->{$field} = $val || $default;
}
sub _set_limiter ($$$) {
- my ($self, $pi_config, $pfx) = @_;
+ my ($self, $pi_cfg, $pfx) = @_;
my $lkey = "-${pfx}_limiter";
$self->{$lkey} ||= do {
# full key is: publicinbox.$NAME.httpbackendmax
require PublicInbox::Qspawn;
$lim = PublicInbox::Qspawn::Limiter->new($val);
} elsif ($val =~ /\A[a-z][a-z0-9]*\z/) {
- $lim = $pi_config->limiter($val);
+ $lim = $pi_cfg->limiter($val);
warn "$mkey limiter=$val not found\n" if !$lim;
} else {
warn "$mkey limiter=$val not understood\n";
my $v = $opts->{address} ||= [ 'public-inbox@example.com' ];
my $p = $opts->{-primary_address} = ref($v) eq 'ARRAY' ? $v->[0] : $v;
$opts->{domain} = ($p =~ /\@(\S+)\z/) ? $1 : 'localhost';
- my $pi_config = delete $opts->{-pi_config};
- _set_limiter($opts, $pi_config, 'httpbackend');
- _set_uint($opts, 'feedmax', 25);
- $opts->{nntpserver} ||= $pi_config->{'publicinbox.nntpserver'};
- my $dir = $opts->{inboxdir};
- if (defined $dir && -f "$dir/inbox.lock") {
- $opts->{version} = 2;
+ my $pi_cfg = delete $opts->{-pi_cfg};
+ _set_limiter($opts, $pi_cfg, 'httpbackend');
+ my $fmax = $opts->{feedmax};
+ if (defined($fmax) && $fmax =~ /\A[0-9]+\z/) {
+ $opts->{feedmax} += 0;
+ } else {
+ delete $opts->{feedmax};
}
-
# allow any combination of multi-line or comma-delimited hide entries
my $hide = {};
if (defined(my $h = $opts->{hide})) {
bless $opts, $class;
}
-sub version { $_[0]->{version} // 1 }
+sub version {
+ $_[0]->{version} //= -f "$_[0]->{inboxdir}/inbox.lock" ? 2 : 1
+}
sub git_epoch {
my ($self, $epoch) = @_; # v2-only, callers always supply $epoch
- $self->{"$epoch.git"} ||= do {
+ $self->{"$epoch.git"} //= do {
my $git_dir = "$self->{inboxdir}/git/$epoch.git";
return unless -d $git_dir;
my $g = PublicInbox::Git->new($git_dir);
- $g->{-httpbackend_limiter} = $self->{-httpbackend_limiter};
- # no cleanup needed, we never cat-file off this, only clone
+ my $lim = $self->{-httpbackend_limiter};
+ $g->{-httpbackend_limiter} = $lim if $lim;
+ # caller must manually cleanup when done
$g;
};
}
sub git {
my ($self) = @_;
- $self->{git} ||= do {
+ $self->{git} //= do {
my $git_dir = $self->{inboxdir};
$git_dir .= '/all.git' if $self->version == 2;
my $g = PublicInbox::Git->new($git_dir);
- $g->{-httpbackend_limiter} = $self->{-httpbackend_limiter};
+ my $lim = $self->{-httpbackend_limiter};
+ $g->{-httpbackend_limiter} = $lim if $lim;
_cleanup_later($self);
$g;
};
my ($self) = @_;
return if $self->version < 2;
my $cur = $self->{-max_git_epoch};
- my $changed = git($self)->alternates_changed;
- if (!defined($cur) || $changed) {
+ my $changed;
+ if (!defined($cur) || ($changed = git($self)->alternates_changed)) {
git_cleanup($self) if $changed;
my $gits = "$self->{inboxdir}/git";
if (opendir my $dh, $gits) {
- my $max = -1;
- while (defined(my $git_dir = readdir($dh))) {
- $git_dir =~ m!\A([0-9]+)\.git\z! or next;
- $max = $1 if $1 > $max;
- }
- $cur = $self->{-max_git_epoch} = $max if $max >= 0;
- } else {
- warn "opendir $gits failed: $!\n";
+ my $max = max(map {
+ substr($_, 0, -4) + 0; # drop ".git" suffix
+ } grep(/\A[0-9]+\.git\z/, readdir($dh))) // return;
+ $cur = $self->{-max_git_epoch} = $max;
}
}
$cur;
}
sub mm {
- my ($self) = @_;
- $self->{mm} ||= eval {
+ my ($self, $req) = @_;
+ $self->{mm} //= eval {
require PublicInbox::Msgmap;
my $dir = $self->{inboxdir};
if ($self->version >= 2) {
} else {
PublicInbox::Msgmap->new($dir);
}
- };
+ } // ($req ? croak("E: $@") : undef);
}
sub search {
my ($self) = @_;
- my $srch = $self->{search} //= eval {
+ $self->{search} // eval {
_cleanup_later($self);
require PublicInbox::Search;
- PublicInbox::Search->new($self);
+ my $srch = PublicInbox::Search->new($self);
+ (eval { $srch->xdb }) ? ($self->{search} = $srch) : undef;
};
- (eval { $srch->xdb }) ? $srch : undef;
}
# isrch is preferred for read-only interfaces if available since it
sub isrch { $_[0]->{isrch} // search($_[0]) }
sub over {
- $_[0]->{over} //= eval {
- my $srch = $_[0]->{search} //= eval {
- _cleanup_later($_[0]);
+ my ($self, $req) = @_;
+ $self->{over} // eval {
+ my $srch = $self->{search} // do {
require PublicInbox::Search;
- PublicInbox::Search->new($_[0]);
+ PublicInbox::Search->new($self);
};
my $over = PublicInbox::Over->new("$srch->{xpfx}/over.sqlite3");
$over->dbh; # may fail
- $over;
- };
+ $self->{over} = $over;
+ } // ($req ? croak("E: $@") : undef);
}
-
sub try_cat {
my ($path) = @_;
- my $rv = '';
- if (open(my $fh, '<', $path)) {
- local $/;
- $rv = <$fh>;
- }
- $rv;
+ open(my $fh, '<', $path) or return '';
+ local $/;
+ <$fh> // '';
}
sub cat_desc ($) {
sub cloneurl {
my ($self) = @_;
- ($self->{cloneurl} //= do {
+ $self->{cloneurl} // do {
my $s = try_cat("$self->{inboxdir}/cloneurl");
my @urls = split(/\s+/s, $s);
- scalar(@urls) ? \@urls : undef
- }) // [];
+ scalar(@urls) ? ($self->{cloneurl} = \@urls) : undef;
+ } // [];
}
sub base_url {
my ($self, $env) = @_; # env - PSGI env
- if ($env) {
+ if ($env && $env->{'psgi.url_scheme'}) {
my $url = PublicInbox::Git::host_prefix_url($env, '');
# for mount in Plack::Builder
$url .= '/' if $url !~ m!/\z!;
}
# called from a non-PSGI environment (e.g. NNTP/POP3):
$self->{-base_url} ||= do {
- my $url = $self->{url}->[0] or return undef;
+ my $url = $self->{url} // return undef;
+ $url = $url->[0] // return undef;
# expand protocol-relative URLs to HTTPS if we're
# not inside a web server
$url = "https:$url" if $url =~ m!\A//!;
};
}
-sub nntp_url {
- my ($self) = @_;
- $self->{-nntp_url} ||= do {
- # no checking for nntp_usable here, we can point entirely
- # to non-local servers or users run by a different user
- my $ns = $self->{nntpserver};
- my $group = $self->{newsgroup};
- my @urls;
- if ($ns && $group) {
- $ns = [ $ns ] if ref($ns) ne 'ARRAY';
- @urls = map {
- my $u = m!\Anntps?://! ? $_ : "nntp://$_";
- $u .= '/' if $u !~ m!/\z!;
- $u.$group;
- } @$ns;
- }
-
- my $mirrors = $self->{nntpmirror};
- if ($mirrors) {
- my @m;
- foreach (@$mirrors) {
- my $u = m!\Anntps?://! ? $_ : "nntp://$_";
- if ($u =~ m!\Anntps?://[^/]+/?\z!) {
- if ($group) {
- $u .= '/' if $u !~ m!/\z!;
- $u .= $group;
- } else {
- warn
-"publicinbox.$self->{name}.nntpmirror=$_ missing newsgroup name\n";
- }
+sub _x_url ($$$) {
+ my ($self, $x, $ctx) = @_; # $x is "nntp" or "imap"
+ # no checking for nntp_usable here, we can point entirely
+ # to non-local servers or users run by a different user
+ my $ns = $self->{"${x}server"} //
+ $ctx->{www}->{pi_cfg}->get_all("publicinbox.${x}server");
+ my $group = $self->{newsgroup};
+ my @urls;
+ if ($ns && $group) {
+ @urls = map {
+ my $u = m!\A${x}s?://! ? $_ : "$x://$_";
+ $u .= '/' if $u !~ m!/\z!;
+ $u.$group;
+ } @$ns;
+ }
+ if (my $mirrors = $self->{"${x}mirror"}) {
+ my @m;
+ for (@$mirrors) {
+ my $u = m!\A${x}s?://! ? $_ : "$x://$_";
+ if ($u =~ m!\A${x}s?://[^/]+/?\z!) {
+ if ($group) {
+ $u .= '/' if $u !~ m!/\z!;
+ $u .= $group;
+ } else { # n.b. IMAP uses "newsgroup"
+ warn <<EOM;
+publicinbox.$self->{name}.${x}mirror=$_ missing newsgroup name
+EOM
}
- # else: allow full URLs like:
- # nntp://news.example.com/alt.example
- push @m, $u;
}
-
- # List::Util::uniq requires Perl 5.26+, maybe we
- # can use it by 2030 or so
- my %seen;
- @urls = grep { !$seen{$_}++ } (@urls, @m);
+ # else: allow full URLs like:
+ # nntp://news.example.com/alt.example
+ push @m, $u;
}
- \@urls;
- };
+
+ # List::Util::uniq requires Perl 5.26+, maybe we
+ # can use it by 2030 or so
+ my %seen;
+ @urls = grep { !$seen{$_}++ } (@urls, @m);
+ }
+ \@urls;
}
+# my ($self, $ctx) = @_;
+sub nntp_url { $_[0]->{-nntp_url} //= _x_url($_[0], 'nntp', $_[1]) }
+sub imap_url { $_[0]->{-imap_url} //= _x_url($_[0], 'imap', $_[1]) }
+
sub nntp_usable {
my ($self) = @_;
my $ret = mm($self) && over($self);
- $self->{mm} = $self->{over} = $self->{search} = undef;
+ delete @$self{qw(mm over search)};
$ret;
}
my ($self) = @_;
check_inodes($self);
my $subs = $self->{unlock_subs} or return;
- for (values %$subs) {
- eval { $_->on_inbox_unlock($self) };
+ for my $obj (values %$subs) {
+ eval { $obj->on_inbox_unlock($self) };
warn "E: $@ ($self->{inboxdir})\n" if $@;
}
}
-sub uidvalidity { $_[0]->{uidvalidity} //= $_[0]->mm->created_at }
+sub uidvalidity { $_[0]->{uidvalidity} //= eval { $_[0]->mm->created_at } }
sub eidx_key { $_[0]->{newsgroup} // $_[0]->{inboxdir} }
+sub mailboxid { # rfc 8474, 8620, 8621
+ my ($self, $imap_slice) = @_;
+ my $pfx = defined($imap_slice) ? $self->{newsgroup} : $self->{name};
+ utf8::encode($pfx); # to octets
+ # RFC 8620, 1.2 recommends not starting with dash or digits
+ # "A good solution to these issues is to prefix every id with a single
+ # alphabetical character."
+ 'M'.join('', map { sprintf('%02x', ord) } split(//, $pfx)) .
+ (defined($imap_slice) ? sprintf('-%x', $imap_slice) : '') .
+ sprintf('-%x', uidvalidity($self) // 0)
+}
+
1;