# slow storage.
#
# data notes:
-# * NNTP article numbers are UIDs and message sequence numbers (MSNs)
-# * Message sequence numbers (MSNs) can be stable since we're read-only.
-# Most IMAP clients use UIDs (I hope). We may return a dummy message
-# in the future if a client requests a non-existent MSN, but that seems
-# unecessary with mutt.
-
+#
+# * NNTP article numbers are UIDs, mm->created_at is UIDVALIDITY
+#
+# * public-inboxes are sliced into mailboxes of 50K messages
+# to not overload MUAs: $NEWSGROUP_NAME.$SLICE_INDEX
+# Slices are similar in concept to v2 "epochs". Epochs
+# are for the limitations of git clients, while slices are
+# for the limitations of IMAP clients.
+#
+# * We also take advantage of slices being only 50K to store
+# "UID offset" to message sequence number (MSN) mapping
+# as a 50K uint16_t array (via pack("S*", ...)). "UID offset"
+# is the offset from {uid_base} which determines the start of
+# the mailbox slice.
+#
+# fields:
+# imapd: PublicInbox::IMAPD ref
+# ibx: PublicInbox::Inbox ref
+# long_cb: long_response private data
+# uid_base: base UID for mailbox slice (0-based)
+# -login_tag: IMAP TAG for LOGIN
+# -idle_tag: IMAP response tag for IDLE
+# uo2m: UID-to-MSN mapping
package PublicInbox::IMAP;
use strict;
-use base qw(PublicInbox::DS);
-use fields qw(imapd logged_in ibx long_cb -login_tag
- uid_min -idle_tag -idle_max);
+use parent qw(PublicInbox::DS);
use PublicInbox::Eml;
use PublicInbox::EmlContentFoo qw(parse_content_disposition);
use PublicInbox::DS qw(now);
use PublicInbox::GitAsyncCat;
use Text::ParseWords qw(parse_line);
use Errno qw(EAGAIN);
-use Time::Local qw(timegm);
-use POSIX qw(strftime);
+use PublicInbox::IMAPsearchqp;
my $Address;
for my $mod (qw(Email::Address::XS Mail::Address)) {
}
die "neither Email::Address::XS nor Mail::Address loaded: $@" if !$Address;
-sub LINE_MAX () { 512 } # does RFC 3501 have a limit like RFC 977?
+sub LINE_MAX () { 8000 } # RFC 2683 3.2.1.5
-# changing this will cause grief for clients which cache
-sub UID_BLOCK () { 50_000 }
+# Changing UID_SLICE will cause grief for clients which cache.
+# This also needs to be <64K: we pack it into a uint16_t
+# for long_response UID (offset) => MSN mappings
+sub UID_SLICE () { 50_000 }
# these values area also used for sorting
sub NEED_SMSG () { 1 }
sub NEED_BLOB () { NEED_SMSG|2 }
-sub NEED_EML () { NEED_BLOB|4 }
-my $OP_EML_NEW = [ NEED_EML - 1, \&op_eml_new ];
+sub CRLF_BREF () { 4 }
+sub EML_HDR () { 8 }
+sub CRLF_HDR () { 16 }
+sub EML_BDY () { 32 }
+sub CRLF_BDY () { 64 }
+my $OP_EML_NEW = [ EML_HDR - 1, \&op_eml_new ];
+my $OP_CRLF_BREF = [ CRLF_BREF, \&op_crlf_bref ];
+my $OP_CRLF_HDR = [ CRLF_HDR, \&op_crlf_hdr ];
+my $OP_CRLF_BDY = [ CRLF_BDY, \&op_crlf_bdy ];
my %FETCH_NEED = (
- 'BODY[HEADER]' => [ NEED_EML, \&emit_rfc822_header ],
- 'BODY[TEXT]' => [ NEED_EML, \&emit_rfc822_text ],
- 'BODY[]' => [ NEED_BLOB, \&emit_rfc822 ],
- 'RFC822.HEADER' => [ NEED_EML, \&emit_rfc822_header ],
- 'RFC822.TEXT' => [ NEED_EML, \&emit_rfc822_text ],
- 'RFC822.SIZE' => [ NEED_BLOB, \&emit_rfc822_size ],
- RFC822 => [ NEED_BLOB, \&emit_rfc822 ],
- BODY => [ NEED_EML, \&emit_body ],
- BODYSTRUCTURE => [ NEED_EML, \&emit_bodystructure ],
- ENVELOPE => [ NEED_EML, \&emit_envelope ],
+ 'BODY[HEADER]' => [ NEED_BLOB|EML_HDR|CRLF_HDR, \&emit_rfc822_header ],
+ 'BODY[TEXT]' => [ NEED_BLOB|EML_BDY|CRLF_BDY, \&emit_rfc822_text ],
+ 'BODY[]' => [ NEED_BLOB|CRLF_BREF, \&emit_rfc822 ],
+ 'RFC822.HEADER' => [ NEED_BLOB|EML_HDR|CRLF_HDR, \&emit_rfc822_header ],
+ 'RFC822.TEXT' => [ NEED_BLOB|EML_BDY|CRLF_BDY, \&emit_rfc822_text ],
+ 'RFC822.SIZE' => [ NEED_SMSG, \&emit_rfc822_size ],
+ RFC822 => [ NEED_BLOB|CRLF_BREF, \&emit_rfc822 ],
+ BODY => [ NEED_BLOB|EML_HDR|EML_BDY, \&emit_body ],
+ BODYSTRUCTURE => [ NEED_BLOB|EML_HDR|EML_BDY, \&emit_bodystructure ],
+ ENVELOPE => [ NEED_BLOB|EML_HDR, \&emit_envelope ],
FLAGS => [ 0, \&emit_flags ],
INTERNALDATE => [ NEED_SMSG, \&emit_internaldate ],
);
my $valid_range = '[0-9]+|[0-9]+:[0-9]+|[0-9]+:\*';
$valid_range = qr/\A(?:$valid_range)(?:,(?:$valid_range))*\z/;
-my @MoY = qw(Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec);
-my %MoY;
-@MoY{@MoY} = (0..11);
+# RFC 3501 5.4. Autologout Timer needs to be >= 30min
+$PublicInbox::DS::EXPTIME = 60 * 30;
sub greet ($) {
my ($self) = @_;
sub new ($$$) {
my ($class, $sock, $imapd) = @_;
- my $self = fields::new($class);
+ my $self = bless { imapd => $imapd }, 'PublicInbox::IMAP_preauth';
my $ev = EPOLLIN;
my $wbuf;
if ($sock->can('accept_SSL') && !$sock->accept_SSL) {
$wbuf = [ \&PublicInbox::DS::accept_tls_step, \&greet ];
}
$self->SUPER::new($sock, $ev | EPOLLONESHOT);
- $self->{imapd} = $imapd;
if ($wbuf) {
$self->{wbuf} = $wbuf;
} else {
$self;
}
+sub logged_in { 1 }
+
sub capa ($) {
my ($self) = @_;
# dovecot advertises IDLE pre-login; perhaps because some clients
# depend on it, so we'll do the same
my $capa = 'CAPABILITY IMAP4rev1 IDLE';
- if ($self->{logged_in}) {
+ if ($self->logged_in) {
$capa .= ' COMPRESS=DEFLATE';
} else {
if (!($self->{sock} // $self)->can('accept_SSL') &&
sub login_success ($$) {
my ($self, $tag) = @_;
- $self->{logged_in} = 1;
+ bless $self, 'PublicInbox::IMAP';
my $capa = capa($self);
"$tag OK [$capa] Logged in\r\n";
}
sub cmd_close ($$) {
my ($self, $tag) = @_;
- delete $self->{uid_min};
+ delete @$self{qw(uid_base uo2m)};
delete $self->{ibx} ? "$tag OK Close done\r\n"
: "$tag BAD No mailbox\r\n";
}
sub cmd_logout ($$) {
my ($self, $tag) = @_;
- delete $self->{logged_in};
+ delete $self->{-idle_tag};
$self->write(\"* BYE logging out\r\n$tag OK Logout done\r\n");
$self->shutdn; # PublicInbox::DS::shutdn
undef;
'* '.capa($self)."\r\n$tag OK Capability done\r\n";
}
-sub cmd_noop ($$) { "$_[1] OK Noop done\r\n" }
+# uo2m: UID Offset to MSN, this is an arrayref by default,
+# but uo2m_hibernate can compact and deduplicate it
+sub uo2m_ary_new ($;$) {
+ my ($self, $exists) = @_;
+ my $base = $self->{uid_base};
+ my $uids = $self->{ibx}->over->uid_range($base + 1, $base + UID_SLICE);
+
+ # convert UIDs to offsets from {base}
+ my @tmp; # [$UID_OFFSET] => $MSN
+ my $msn = 0;
+ ++$base;
+ $tmp[$_ - $base] = ++$msn for @$uids;
+ $$exists = $msn if $exists;
+ \@tmp;
+}
+
+# changes UID-offset-to-MSN mapping into a deduplicated scalar:
+# uint16_t uo2m[UID_SLICE].
+# May be swapped out for idle clients if THP is disabled.
+sub uo2m_hibernate ($) {
+ my ($self) = @_;
+ ref(my $uo2m = $self->{uo2m}) or return;
+ my %dedupe = ( uo2m_pack($uo2m) => undef );
+ $self->{uo2m} = (keys(%dedupe))[0];
+ undef;
+}
+
+sub uo2m_last_uid ($) {
+ my ($self) = @_;
+ defined(my $uo2m = $self->{uo2m}) or die 'BUG: uo2m_last_uid w/o {uo2m}';
+ (ref($uo2m) ? @$uo2m : (length($uo2m) >> 1)) + $self->{uid_base};
+}
+
+sub uo2m_pack ($) {
+ # $_[0] is an arrayref of MSNs, it may have undef gaps if there
+ # are gaps in the corresponding UIDs: [ msn1, msn2, undef, msn3 ]
+ no warnings 'uninitialized';
+ pack('S*', @{$_[0]});
+}
+
+# extend {uo2m} to account for new messages which arrived since
+# {uo2m} was created.
+sub uo2m_extend ($$;$) {
+ my ($self, $new_uid_max) = @_;
+ defined(my $uo2m = $self->{uo2m}) or
+ return($self->{uo2m} = uo2m_ary_new($self));
+ my $beg = uo2m_last_uid($self); # last UID we've learned
+ return $uo2m if $beg >= $new_uid_max; # fast path
+
+ # need to extend the current range:
+ my $base = $self->{uid_base};
+ ++$beg;
+ my $uids = $self->{ibx}->over->uid_range($beg, $base + UID_SLICE);
+ return $uo2m if !scalar(@$uids);
+ my @tmp; # [$UID_OFFSET] => $MSN
+ my $write_method = $_[2] // 'msg_more';
+ if (ref($uo2m)) {
+ my $msn = $uo2m->[-1];
+ $tmp[$_ - $beg] = ++$msn for @$uids;
+ $self->$write_method("* $msn EXISTS\r\n");
+ push @$uo2m, @tmp;
+ $uo2m;
+ } else {
+ my $msn = unpack('S', substr($uo2m, -2, 2));
+ $tmp[$_ - $beg] = ++$msn for @$uids;
+ $self->$write_method("* $msn EXISTS\r\n");
+ $uo2m .= uo2m_pack(\@tmp);
+ my %dedupe = ($uo2m => undef);
+ $self->{uo2m} = (keys %dedupe)[0];
+ }
+}
+
+sub cmd_noop ($$) {
+ my ($self, $tag) = @_;
+ defined($self->{uid_base}) and
+ uo2m_extend($self, $self->{uid_base} + UID_SLICE);
+ \"$tag OK Noop done\r\n";
+}
+
+# the flexible version which works on scalars and array refs.
+# Must call uo2m_extend before this
+sub uid2msn ($$) {
+ my ($self, $uid) = @_;
+ my $uo2m = $self->{uo2m};
+ my $off = $uid - $self->{uid_base} - 1;
+ ref($uo2m) ? $uo2m->[$off] : unpack('S', substr($uo2m, $off << 1, 2));
+}
+
+# returns an arrayref of UIDs, so MSNs can be translated to UIDs via:
+# $msn2uid->[$MSN-1] => $UID. The result of this is always ephemeral
+# and does not live beyond the event loop.
+sub msn2uid ($) {
+ my ($self) = @_;
+ my $base = $self->{uid_base};
+ my $uo2m = uo2m_extend($self, $base + UID_SLICE);
+ $uo2m = [ unpack('S*', $uo2m) ] if !ref($uo2m);
+
+ my $uo = 0;
+ my @msn2uid;
+ for my $msn (@$uo2m) {
+ ++$uo;
+ $msn2uid[$msn - 1] = $uo + $base if $msn;
+ }
+ \@msn2uid;
+}
+
+# converts a set of message sequence numbers in requests to UIDs:
+sub msn_to_uid_range ($$) {
+ my $msn2uid = $_[0];
+ $_[1] =~ s!([0-9]+)!$msn2uid->[$1 - 1] // ($msn2uid->[-1] + 1)!sge;
+}
# called by PublicInbox::InboxIdle
sub on_inbox_unlock {
my ($self, $ibx) = @_;
- my $new = $ibx->mm->max;
- defined(my $old = $self->{-idle_max}) or die 'BUG: -idle_max unset';
- if ($new > $old) {
- $self->{-idle_max} = $new;
- $self->msg_more("* $_ EXISTS\r\n") for (($old + 1)..($new - 1));
- $self->write(\"* $new EXISTS\r\n");
+ my $uid_end = $self->{uid_base} + UID_SLICE;
+ uo2m_extend($self, $uid_end, 'write');
+ my $new = uo2m_last_uid($self);
+ if ($new == $uid_end) { # max exceeded $uid_end
+ # continue idling w/o inotify
+ my $sock = $self->{sock} or return;
+ $ibx->unsubscribe_unlock(fileno($sock));
+ }
+}
+
+# called every X minute(s) or so by PublicInbox::DS::later
+my $IDLERS = {};
+my $idle_timer;
+sub idle_tick_all {
+ my $old = $IDLERS;
+ $IDLERS = {};
+ for my $i (values %$old) {
+ next if ($i->{wbuf} || !exists($i->{-idle_tag}));
+ $i->update_idle_time or next;
+ $IDLERS->{fileno($i->{sock})} = $i;
+ $i->write(\"* OK Still here\r\n");
}
+ $idle_timer = scalar keys %$IDLERS ?
+ PublicInbox::DS::later(\&idle_tick_all) : undef;
}
sub cmd_idle ($$) {
my ($self, $tag) = @_;
# IDLE seems allowed by dovecot w/o a mailbox selected *shrug*
my $ibx = $self->{ibx} or return "$tag BAD no mailbox selected\r\n";
- $ibx->subscribe_unlock(fileno($self->{sock}), $self);
- $self->{imapd}->idler_start;
+ my $uid_end = $self->{uid_base} + UID_SLICE;
+ uo2m_extend($self, $uid_end);
+ my $sock = $self->{sock} or return;
+ my $fd = fileno($sock);
$self->{-idle_tag} = $tag;
- $self->{-idle_max} = $ibx->mm->max // 0;
- "+ idling\r\n"
+ # only do inotify on most recent slice
+ if ($ibx->over->max < $uid_end) {
+ $ibx->subscribe_unlock($fd, $self);
+ $self->{imapd}->idler_start;
+ }
+ $idle_timer //= PublicInbox::DS::later(\&idle_tick_all);
+ $IDLERS->{$fd} = $self;
+ \"+ idling\r\n"
}
-sub cmd_done ($$) {
+sub stop_idle ($$) {
+ my ($self, $ibx) = @_;
+ my $sock = $self->{sock} or return;
+ my $fd = fileno($sock);
+ delete $IDLERS->{$fd};
+ $ibx->unsubscribe_unlock($fd);
+}
+
+sub idle_done ($$) {
my ($self, $tag) = @_; # $tag is "DONE" (case-insensitive)
defined(my $idle_tag = delete $self->{-idle_tag}) or
return "$tag BAD not idle\r\n";
warn "BUG: idle_tag set w/o inbox";
return "$tag BAD internal bug\r\n";
};
- $ibx->unsubscribe_unlock(fileno($self->{sock}));
+ stop_idle($self, $ibx);
"$idle_tag OK Idle done\r\n";
}
-sub ensure_ranges_exist ($$$) {
+sub ensure_slices_exist ($$$) {
my ($imapd, $ibx, $max) = @_;
+ defined(my $mb_top = $ibx->{newsgroup}) or return;
my $mailboxes = $imapd->{mailboxes};
- my $mb_top = $ibx->{newsgroup};
my @created;
- for (my $i = int($max/UID_BLOCK); $i >= 0; --$i) {
+ for (my $i = int($max/UID_SLICE); $i >= 0; --$i) {
my $sub_mailbox = "$mb_top.$i";
last if exists $mailboxes->{$sub_mailbox};
$mailboxes->{$sub_mailbox} = $ibx;
+ $sub_mailbox =~ s/\Ainbox\./INBOX./i; # more familiar to users
push @created, $sub_mailbox;
}
return unless @created;
push @$l, map { qq[* LIST (\\HasNoChildren) "." $_\r\n] } @created;
}
-sub inbox_lookup ($$) {
- my ($self, $mailbox) = @_;
- my ($ibx, $exists, $uidnext);
- if ($mailbox =~ /\A(.+)\.([0-9]+)\z/) {
- # old mail: inbox.comp.foo.$uid_block_idx
- my ($mb_top, $uid_min) = ($1, $2 * UID_BLOCK + 1);
-
- $ibx = $self->{imapd}->{mailboxes}->{lc $mailbox} or return;
- $exists = $ibx->mm->max // 0;
- $self->{uid_min} = $uid_min;
- ensure_ranges_exist($self->{imapd}, $ibx, $exists);
- my $uid_end = $uid_min + UID_BLOCK - 1;
- $exists = $uid_end if $exists > $uid_end;
- $uidnext = $exists + 1;
- } else { # check for dummy inboxes
- $ibx = $self->{imapd}->{mailboxes}->{lc $mailbox} or return;
- delete $self->{uid_min};
- $exists = 0;
- $uidnext = 1;
+sub inbox_lookup ($$;$) {
+ my ($self, $mailbox, $examine) = @_;
+ my ($ibx, $exists, $uidmax, $uid_base) = (undef, 0, 0, 0);
+ $mailbox = lc $mailbox;
+ $ibx = $self->{imapd}->{mailboxes}->{$mailbox} or return;
+ my $over = $ibx->over;
+ if ($over != $ibx) { # not a dummy
+ $mailbox =~ /\.([0-9]+)\z/ or
+ die "BUG: unexpected dummy mailbox: $mailbox\n";
+ $uid_base = $1 * UID_SLICE;
+
+ # ->num_highwater caches for writers, so use ->meta_accessor
+ $uidmax = $ibx->mm->meta_accessor('num_highwater') // 0;
+ if ($examine) {
+ $self->{uid_base} = $uid_base;
+ $self->{ibx} = $ibx;
+ $self->{uo2m} = uo2m_ary_new($self, \$exists);
+ } else {
+ $exists = $over->imap_exists;
+ }
+ ensure_slices_exist($self->{imapd}, $ibx, $over->max);
+ } else {
+ if ($examine) {
+ $self->{uid_base} = $uid_base;
+ $self->{ibx} = $ibx;
+ delete $self->{uo2m};
+ }
+ # if "INBOX.foo.bar" is selected and "INBOX.foo.bar.0",
+ # check for new UID ranges (e.g. "INBOX.foo.bar.1")
+ if (my $z = $self->{imapd}->{mailboxes}->{"$mailbox.0"}) {
+ ensure_slices_exist($self->{imapd}, $z, $z->over->max);
+ }
}
- ($ibx, $exists, $uidnext);
+ ($ibx, $exists, $uidmax + 1, $uid_base);
}
sub cmd_examine ($$$) {
my ($self, $tag, $mailbox) = @_;
- my ($ibx, $exists, $uidnext) = inbox_lookup($self, $mailbox);
- return "$tag NO Mailbox doesn't exist: $mailbox\r\n" if !$ibx;
-
# XXX: do we need this? RFC 5162/7162
my $ret = $self->{ibx} ? "* OK [CLOSED] previous closed\r\n" : '';
- $self->{ibx} = $ibx;
+ my ($ibx, $exists, $uidnext, $base) = inbox_lookup($self, $mailbox, 1);
+ return "$tag NO Mailbox doesn't exist: $mailbox\r\n" if !$ibx;
$ret .= <<EOF;
* $exists EXISTS\r
* $exists RECENT\r
$self->requeue if $new_size == 1;
}
-sub uid_fetch_cb { # called by git->cat_async via git_async_cat
- my ($bref, $oid, $type, $size, $fetch_m_arg) = @_;
- my ($self, undef, $msgs, undef, $ops, $partial) = @$fetch_m_arg;
+sub fetch_run_ops {
+ my ($self, $smsg, $bref, $ops, $partial) = @_;
+ my $uid = $smsg->{num};
+ $self->msg_more('* '.uid2msn($self, $uid)." FETCH (UID $uid");
+ my ($eml, $k);
+ for (my $i = 0; $i < @$ops;) {
+ $k = $ops->[$i++];
+ $ops->[$i++]->($self, $k, $smsg, $bref, $eml);
+ }
+ partial_emit($self, $partial, $eml) if $partial;
+ $self->msg_more(")\r\n");
+}
+
+sub fetch_blob_cb { # called by git->cat_async via git_async_cat
+ my ($bref, $oid, $type, $size, $fetch_arg) = @_;
+ my ($self, undef, $msgs, $range_info, $ops, $partial) = @$fetch_arg;
my $smsg = shift @$msgs or die 'BUG: no smsg';
if (!defined($oid)) {
# it's possible to have TOCTOU if an admin runs
# public-inbox-(edit|purge), just move onto the next message
+ warn "E: $smsg->{blob} missing in $self->{ibx}->{inboxdir}\n";
return requeue_once($self);
} else {
$smsg->{blob} eq $oid or die "BUG: $smsg->{blob} != $oid";
}
- $$bref =~ s/(?<!\r)\n/\r\n/sg; # make strict clients happy
-
- # fixup old bug from import (pre-a0c07cba0e5d8b6a)
- $$bref =~ s/\A[\r\n]*From [^\r\n]*\r\n//s;
- $self->msg_more("* $smsg->{num} FETCH (UID $smsg->{num}");
- my $eml;
- for (my $i = 0; $i < @$ops;) {
- my $k = $ops->[$i++];
- $ops->[$i++]->($self, $k, $smsg, $bref, $eml);
+ my $pre;
+ if (!$self->{wbuf} && (my $nxt = $msgs->[0])) {
+ $pre = $self->{ibx}->git->async_prefetch($nxt->{blob},
+ \&fetch_blob_cb, $fetch_arg);
}
- partial_emit($self, $partial, $eml) if $partial;
- $self->msg_more(")\r\n");
- requeue_once($self);
+ fetch_run_ops($self, $smsg, $bref, $ops, $partial);
+ $pre ? $self->zflush : requeue_once($self);
}
sub emit_rfc822 {
$self->msg_more($$bref);
}
-# Mail::IMAPClient::message_string cares about this by default
-# (->Ignoresizeerrors attribute)
+# Mail::IMAPClient::message_string cares about this by default,
+# (->Ignoresizeerrors attribute). Admins are encouraged to
+# --reindex for IMAP support, anyways.
sub emit_rfc822_size {
- my ($self, $k, undef, $bref) = @_;
- $self->msg_more(' RFC822.SIZE ' . length($$bref));
+ my ($self, $k, $smsg) = @_;
+ $self->msg_more(' RFC822.SIZE ' . $smsg->{bytes});
}
sub emit_internaldate {
# set $eml once ($_[4] == $eml, $_[3] == $bref)
sub op_eml_new { $_[4] = PublicInbox::Eml->new($_[3]) }
+# s/From / fixes old bug from import (pre-a0c07cba0e5d8b6a)
+sub to_crlf_full {
+ ${$_[0]} =~ s/(?<!\r)\n/\r\n/sg;
+ ${$_[0]} =~ s/\A[\r\n]*From [^\r\n]*\r\n//s;
+}
+
+sub op_crlf_bref { to_crlf_full($_[3]) }
+
+sub op_crlf_hdr { to_crlf_full($_[4]->{hdr}) }
+
+sub op_crlf_bdy { ${$_[4]->{bdy}} =~ s/(?<!\r)\n/\r\n/sg if $_[4]->{bdy} }
+
sub uid_clamp ($$$) {
my ($self, $beg, $end) = @_;
- my $uid_min = $self->{uid_min} or return;
- my $uid_end = $uid_min + UID_BLOCK - 1;
+ my $uid_min = $self->{uid_base} + 1;
+ my $uid_end = $uid_min + UID_SLICE - 1;
$$beg = $uid_min if $$beg < $uid_min;
$$end = $uid_end if $$end > $uid_end;
}
$range = $$range_csv;
$$range_csv = undef;
}
+ my $uid_base = $self->{uid_base};
+ my $uid_end = $uid_base + UID_SLICE;
if ($range =~ /\A([0-9]+):([0-9]+)\z/) {
($beg, $end) = ($1 + 0, $2 + 0);
+ uid_clamp($self, \$beg, \$end);
} elsif ($range =~ /\A([0-9]+):\*\z/) {
$beg = $1 + 0;
- $end = $self->{ibx}->mm->max // 0;
- my $uid_end = ($self->{uid_min} // 1) - 1 + UID_BLOCK;
+ $end = $self->{ibx}->over->max;
$end = $uid_end if $end > $uid_end;
$beg = $end if $beg > $end;
+ uid_clamp($self, \$beg, \$end);
} elsif ($range =~ /\A[0-9]+\z/) {
$beg = $end = $range + 0;
- undef $range;
+ # just let the caller do an out-of-range query if a single
+ # UID is out-of-range
+ ++$beg if ($beg <= $uid_base || $end > $uid_end);
} else {
return 'BAD fetch range';
}
- uid_clamp($self, \$beg, \$end) if defined($range);
[ $beg, $end, $$range_csv ];
}
undef; # keep looping
}
-sub uid_fetch_msg { # long_response
- my ($self, $tag, $msgs, $range_info) = @_; # \@ops, \@partial
+sub fetch_blob { # long_response
+ my ($self, $tag, $msgs, $range_info, $ops, $partial) = @_;
while (!@$msgs) { # rare
if (my $end = refill_range($self, $msgs, $range_info)) {
$self->write(\"$tag $end\r\n");
return;
}
}
+ uo2m_extend($self, $msgs->[-1]->{num});
git_async_cat($self->{ibx}->git, $msgs->[0]->{blob},
- \&uid_fetch_cb, \@_);
+ \&fetch_blob_cb, \@_);
}
-sub uid_fetch_smsg { # long_response
+sub fetch_smsg { # long_response
my ($self, $tag, $msgs, $range_info, $ops) = @_;
while (!@$msgs) { # rare
if (my $end = refill_range($self, $msgs, $range_info)) {
return;
}
}
- for my $smsg (@$msgs) {
- $self->msg_more("* $smsg->{num} FETCH (UID $smsg->{num}");
- for (my $i = 0; $i < @$ops;) {
- my $k = $ops->[$i++];
- $ops->[$i++]->($self, $k, $smsg);
- }
- $self->msg_more(")\r\n");
- }
+ uo2m_extend($self, $msgs->[-1]->{num});
+ fetch_run_ops($self, $_, undef, $ops) for @$msgs;
@$msgs = ();
1; # more
}
-sub uid_fetch_uid { # long_response
- my ($self, $tag, $uids, $range_info, $ops) = @_;
- while (!@$uids) { # rare
- my ($beg, $end, $range_csv) = @$range_info;
- if (scalar(@$uids = @{$self->{ibx}->over->
- uid_range($beg, $end)})) {
- $range_info->[0] = $uids->[-1] + 1;
- } elsif (!$range_csv) {
- $self->write(\"$tag OK Fetch done\r\n");
+sub refill_uids ($$$;$) {
+ my ($self, $uids, $range_info, $sql) = @_;
+ my ($beg, $end, $range_csv) = @$range_info;
+ my $over = $self->{ibx}->over;
+ while (1) {
+ if (scalar(@$uids = @{$over->uid_range($beg, $end, $sql)})) {
+ $range_info->[0] = $uids->[-1] + 1; # update $beg
return;
+ } elsif (!$range_csv) {
+ return 0;
} else {
my $next_range = range_step($self, \$range_csv);
- if (!ref($next_range)) { # error
- $self->write(\"$tag $next_range\r\n");
- return;
- }
- @$range_info = @$next_range;
+ return $next_range if !ref($next_range); # error
+ ($beg, $end, $range_csv) = @$range_info = @$next_range;
+ # continue looping
}
- # continue looping
}
+}
+
+sub fetch_uid { # long_response
+ my ($self, $tag, $uids, $range_info, $ops) = @_;
+ if (defined(my $err = refill_uids($self, $uids, $range_info))) {
+ $err ||= 'OK Fetch done';
+ $self->write("$tag $err\r\n");
+ return;
+ }
+ my $adj = $self->{uid_base} + 1;
+ my $uo2m = uo2m_extend($self, $uids->[-1]);
+ $uo2m = [ unpack('S*', $uo2m) ] if !ref($uo2m);
+ my ($i, $k);
for (@$uids) {
- $self->msg_more("* $_ FETCH (UID $_");
- for (my $i = 0; $i < @$ops;) {
- my $k = $ops->[$i++];
+ $self->msg_more("* $uo2m->[$_ - $adj] FETCH (UID $_");
+ for ($i = 0; $i < @$ops;) {
+ $k = $ops->[$i++];
$ops->[$i++]->($self, $k);
}
$self->msg_more(")\r\n");
# request for hierarchy delimiter
$l = [ qq[* LIST (\\Noselect) "." ""\r\n] ];
} elsif ($refname ne '' || $wildcard ne '*') {
- $wildcard = lc $wildcard;
- $wildcard =~ s!([^a-z0-9_])!$patmap{$1} // "\Q$1"!eg;
- $l = [ grep(/ \Q$refname\E$wildcard\r\n\z/s, @$l) ];
+ $wildcard =~ s!([^a-z0-9_])!$patmap{$1} // "\Q$1"!egi;
+ $l = [ grep(/ \Q$refname\E$wildcard\r\n\z/is, @$l) ];
}
\(join('', @$l, "$tag OK List done\r\n"));
}
}
my $str = $eml->header_obj->as_string;
$str =~ s/$hdrs_re//g;
+ $str =~ s/(?<!\r)\n/\r\n/sg;
$str .= "\r\n";
}
$eml = eml_body_idx($eml, $section_idx) or return;
}
my $str = $eml->header_obj->as_string;
- join('', ($str =~ m/($hdrs_re)/g), "\r\n");
+ $str = join('', ($str =~ m/($hdrs_re)/g));
+ $str =~ s/(?<!\r)\n/\r\n/sg;
+ $str .= "\r\n";
}
-sub partial_prepare ($$$) {
- my ($partial, $want, $att) = @_;
+sub partial_prepare ($$$$) {
+ my ($need, $partial, $want, $att) = @_;
# recombine [ "BODY[1.HEADER.FIELDS", "(foo", "bar)]" ]
# back to: "BODY[1.HEADER.FIELDS (foo bar)]"
(?:\.(HEADER|MIME|TEXT))? # 2 - section_name
\](?:<([0-9]+)(?:\.([0-9]+))?>)?\z/sx) { # 3, 4
$partial->{$att} = [ \&partial_body, $1, $2, $3, $4 ];
+ $$need |= CRLF_BREF|EML_HDR|EML_BDY;
} elsif ($att =~ /\ABODY\[(?:([0-9]+(?:\.[0-9]+)*)\.)? # 1 - section_idx
(?:HEADER\.FIELDS(\.NOT)?)\x20 # 2
\(([A-Z0-9\-\x20]+)\) # 3 - hdrs
: \&partial_hdr_get,
$1, undef, $4, $5 ];
$tmp->[2] = hdrs_regexp($3);
+
+ # don't emit CRLF_HDR instruction, here, partial_hdr_*
+ # will do CRLF conversion with only the extracted result
+ # and not waste time converting lines we don't care about.
+ $$need |= EML_HDR;
} else {
undef;
}
while (my ($k, $fl_cb) = each %$x) {
next if $seen{$k}++;
$need |= $fl_cb->[0];
-
- # insert a special op to convert $bref to $eml
- # the first time we need it
- if ($need == NEED_EML && !$seen{$need}++) {
- push @op, $OP_EML_NEW;
- }
- # $fl_cb = [ flags, \&emit_foo ]
- push @op, [ @$fl_cb , $k ];
+ push @op, [ @$fl_cb, $k ];
}
- } elsif (!partial_prepare(\%partial, $want, $att)) {
+ } elsif (!partial_prepare(\$need, \%partial, $want, $att)) {
return "BAD param: $att";
}
}
# stabilize partial order for consistency and ease-of-debugging:
if (scalar keys %partial) {
- $need = NEED_EML;
- push @op, $OP_EML_NEW if !$seen{$need}++;
+ $need |= NEED_BLOB;
$r[2] = [ map { [ $_, @{$partial{$_}} ] } sort keys %partial ];
}
- $r[0] = $need & NEED_BLOB ? \&uid_fetch_msg :
- ($need & NEED_SMSG ? \&uid_fetch_smsg : \&uid_fetch_uid);
+ push @op, $OP_EML_NEW if ($need & (EML_HDR|EML_BDY));
+
+ # do we need CRLF conversion?
+ if ($need & CRLF_BREF) {
+ push @op, $OP_CRLF_BREF;
+ } elsif (my $crlf = ($need & (CRLF_HDR|CRLF_BDY))) {
+ if ($crlf == (CRLF_HDR|CRLF_BDY)) {
+ push @op, $OP_CRLF_BREF;
+ } elsif ($need & CRLF_HDR) {
+ push @op, $OP_CRLF_HDR;
+ } else {
+ push @op, $OP_CRLF_BDY;
+ }
+ }
+
+ $r[0] = $need & NEED_BLOB ? \&fetch_blob :
+ ($need & NEED_SMSG ? \&fetch_smsg : \&fetch_uid);
# r[1] = [ $key1, $cb1, $key2, $cb2, ... ]
use sort 'stable'; # makes output more consistent
@r;
}
-sub cmd_uid_fetch ($$$;@) {
+sub cmd_uid_fetch ($$$$;@) {
my ($self, $tag, $range_csv, @want) = @_;
my $ibx = $self->{ibx} or return "$tag BAD No mailbox selected\r\n";
my ($cb, $ops, $partial) = fetch_compile(\@want);
return "$tag $cb\r\n" unless $ops;
+ # cb is one of fetch_blob, fetch_smsg, fetch_uid
$range_csv = 'bad' if $range_csv !~ $valid_range;
my $range_info = range_step($self, \$range_csv);
return "$tag $range_info\r\n" if !ref($range_info);
+ uo2m_hibernate($self) if $cb == \&fetch_blob; # slow, save RAM
long_response($self, $cb, $tag, [], $range_info, $ops, $partial);
}
-sub parse_date ($) { # 02-Oct-1993
- my ($date_text) = @_;
- my ($dd, $mon, $yyyy) = split(/-/, $_[0], 3);
- defined($yyyy) or return;
- my $mm = $MoY{$mon} // return;
- $dd =~ /\A[0123]?[0-9]\z/ or return;
- $yyyy =~ /\A[0-9]{4,}\z/ or return; # Y10K-compatible!
- timegm(0, 0, 0, $dd, $mm, $yyyy);
+sub cmd_fetch ($$$$;@) {
+ my ($self, $tag, $range_csv, @want) = @_;
+ my $ibx = $self->{ibx} or return "$tag BAD No mailbox selected\r\n";
+ my ($cb, $ops, $partial) = fetch_compile(\@want);
+ return "$tag $cb\r\n" unless $ops;
+
+ # cb is one of fetch_blob, fetch_smsg, fetch_uid
+ $range_csv = 'bad' if $range_csv !~ $valid_range;
+ msn_to_uid_range(msn2uid($self), $range_csv);
+ my $range_info = range_step($self, \$range_csv);
+ return "$tag $range_info\r\n" if !ref($range_info);
+ uo2m_hibernate($self) if $cb == \&fetch_blob; # slow, save RAM
+ long_response($self, $cb, $tag, [], $range_info, $ops, $partial);
}
-sub uid_search_uid_range { # long_response
- my ($self, $tag, $beg, $end, $sql) = @_;
- my $uids = $self->{ibx}->over->uid_range($$beg, $end, $sql);
- if (@$uids) {
- $$beg = $uids->[-1] + 1;
- $self->msg_more(join(' ', '', @$uids));
- } else {
- $self->write(\"\r\n$tag OK Search done\r\n");
- undef;
+sub msn_convert ($$) {
+ my ($self, $uids) = @_;
+ my $adj = $self->{uid_base} + 1;
+ my $uo2m = uo2m_extend($self, $uids->[-1]);
+ $uo2m = [ unpack('S*', $uo2m) ] if !ref($uo2m);
+ $_ = $uo2m->[$_ - $adj] for @$uids;
+}
+
+sub search_uid_range { # long_response
+ my ($self, $tag, $sql, $range_info, $want_msn) = @_;
+ my $uids = [];
+ if (defined(my $err = refill_uids($self, $uids, $range_info, $sql))) {
+ $err ||= 'OK Search done';
+ $self->write("\r\n$tag $err\r\n");
+ return;
}
+ msn_convert($self, $uids) if $want_msn;
+ $self->msg_more(join(' ', '', @$uids));
+ 1; # more
}
sub date_search {
# KEYWORD # TODO ? dfpre,dfpost,...
);
-sub parse_query {
- my ($self, $rest) = @_;
- if (uc($rest->[0]) eq 'CHARSET') {
- shift @$rest;
- defined(my $c = shift @$rest) or return 'BAD missing charset';
- $c =~ /\A(?:UTF-8|US-ASCII)\z/ or return 'NO [BADCHARSET]';
- }
+# IMAP allows searching arbitrary headers via "HEADER $HDR_NAME $HDR_VAL"
+# which gets silly expensive. We only allow the headers we already index.
+my %H2X = (%I2X, 'MESSAGE-ID' => 'm:', 'LIST-ID' => 'l:');
- my $sql = ''; # date conditions, {sql} deleted if Xapian is needed
- my $q = { xap => '', sql => \$sql };
- while (@$rest) {
- my $k = uc(shift @$rest);
- # default criteria
- next if $k =~ /\A(?:ALL|RECENT|UNSEEN|NEW)\z/;
- next if $k eq 'AND'; # the default, until we support OR
- if ($k =~ $valid_range) { # sequence numbers == UIDs
- push @{$q->{uid}}, $k;
- } elsif ($k eq 'UID') {
- $k = shift(@$rest) // '';
- $k =~ $valid_range or return 'BAD UID range';
- push @{$q->{uid}}, $k;
- } elsif ($k =~ /\A(?:SENT)?(?:SINCE|ON|BEFORE)\z/) {
- my $d = parse_date(shift(@$rest) // '');
- defined $d or return "BAD $k date format";
- date_search($q, $k, $d);
- } elsif ($k =~ /\A(?:SMALLER|LARGER)\z/) {
- delete $q->{sql}; # can't use over.sqlite3
- my $bytes = shift(@$rest) // '';
- $bytes =~ /\A[0-9]+\z/ or return "BAD $k not a number";
- $q->{xap} .= ' bytes:' . ($k eq 'SMALLER' ?
- '..'.(--$bytes) :
- (++$bytes).'..');
- } elsif (defined(my $xk = $I2X{$k})) {
- delete $q->{sql}; # can't use over.sqlite3
- my $arg = shift @$rest;
- defined($arg) or return "BAD $k no arg";
-
- # Xapian can't handle [*"] in probabilistic terms
- $arg =~ tr/*"//d;
- $q->{xap} .= qq[ $xk:"$arg"];
- } else {
- # TODO: parentheses, OR, NOT ...
- return "BAD $k not supported (yet?)";
- }
+sub xap_append ($$$$) {
+ my ($q, $rest, $k, $xk) = @_;
+ delete $q->{sql}; # can't use over.sqlite3
+ defined(my $arg = shift @$rest) or return "BAD $k no arg";
+
+ # AFAIK Xapian can't handle [*"] in probabilistic terms
+ $arg =~ tr/*"//d;
+ ${$q->{xap}} .= qq[ $xk"$arg"];
+ undef;
+}
+
+sub parse_query ($$) {
+ my ($self, $query) = @_;
+ my $q = PublicInbox::IMAPsearchqp::parse($self, $query);
+ if (ref($q)) {
+ my $max = $self->{ibx}->over->max;
+ my $beg = 1;
+ uid_clamp($self, \$beg, \$max);
+ $q->{range_info} = [ $beg, $max ];
}
+ $q;
+}
- # favor using over.sqlite3 if possible, since Xapian is optional
- if (exists $q->{sql}) {
- delete($q->{xap});
- delete($q->{sql}) if $sql eq '';
- } elsif (!$self->{ibx}->search) {
- return 'BAD Xapian not configured for mailbox';
+sub refill_xap ($$$$) {
+ my ($self, $uids, $range_info, $q) = @_;
+ my ($beg, $end) = @$range_info;
+ my $srch = $self->{ibx}->search;
+ my $opt = { mset => 2, limit => 1000 };
+ my $mset = $srch->query("$q uid:$beg..$end", $opt);
+ @$uids = @{$srch->mset_to_artnums($mset)};
+ if (@$uids) {
+ $range_info->[0] = $uids->[-1] + 1; # update $beg
+ return; # possibly more
}
+ 0; # all done
+}
- if (my $uid = $q->{uid}) {
- ((@$uid > 1) || $uid->[0] =~ /,/) and
- return 'BAD multiple ranges not supported, yet';
- ($q->{sql} // $q->{xap}) and
- return 'BAD ranges and queries do not mix, yet';
- $q->{uid} = join(',', @$uid); # TODO: multiple ranges
+sub search_xap_range { # long_response
+ my ($self, $tag, $q, $range_info, $want_msn) = @_;
+ my $uids = [];
+ if (defined(my $err = refill_xap($self, $uids, $range_info, $q))) {
+ $err ||= 'OK Search done';
+ $self->write("\r\n$tag $err\r\n");
+ return;
}
- $q;
+ msn_convert($self, $uids) if $want_msn;
+ $self->msg_more(join(' ', '', @$uids));
+ 1; # more
}
-sub cmd_uid_search ($$$;) {
- my ($self, $tag) = splice(@_, 0, 2);
+sub search_common {
+ my ($self, $tag, $query, $want_msn) = @_;
my $ibx = $self->{ibx} or return "$tag BAD No mailbox selected\r\n";
- my $q = parse_query($self, \@_);
+ my $q = parse_query($self, $query);
return "$tag $q\r\n" if !ref($q);
- my $sql = delete $q->{sql};
-
- if (!scalar(keys %$q)) {
+ my ($sql, $range_info) = delete @$q{qw(sql range_info)};
+ if (!scalar(keys %$q)) { # overview.sqlite3
$self->msg_more('* SEARCH');
- my $beg = $self->{uid_min} // 1;
- my $end = $ibx->mm->max;
- uid_clamp($self, \$beg, \$end);
- long_response($self, \&uid_search_uid_range,
- $tag, \$beg, $end, $sql);
- } elsif (my $uid = $q->{uid}) {
- if ($uid =~ /\A([0-9]+):([0-9]+|\*)\z/s) {
- my ($beg, $end) = ($1, $2);
- $end = $ibx->mm->max if $end eq '*';
- uid_clamp($self, \$beg, \$end);
- $self->msg_more('* SEARCH');
- long_response($self, \&uid_search_uid_range,
- $tag, \$beg, $end, $sql);
- } elsif ($uid =~ /\A[0-9]+\z/s) {
- $uid = $ibx->over->get_art($uid) ? " $uid" : '';
- "* SEARCH$uid\r\n$tag OK Search done\r\n";
- } else {
- "$tag BAD Error\r\n";
- }
+ long_response($self, \&search_uid_range,
+ $tag, $sql, $range_info, $want_msn);
+ } elsif ($q = $q->{xap}) {
+ $self->{ibx}->search or
+ return "$tag BAD search not available for mailbox\r\n";
+ $self->msg_more('* SEARCH');
+ long_response($self, \&search_xap_range,
+ $tag, $q, $range_info, $want_msn);
} else {
"$tag BAD Error\r\n";
}
}
+sub cmd_uid_search ($$$) {
+ my ($self, $tag, $query) = @_;
+ search_common($self, $tag, $query);
+}
+
+sub cmd_search ($$$;) {
+ my ($self, $tag, $query) = @_;
+ search_common($self, $tag, $query, 1);
+}
+
sub args_ok ($$) { # duplicated from PublicInbox::NNTP
my ($cb, $argc) = @_;
my $tot = prototype $cb;
# returns 1 if we can continue, 0 if not due to buffered writes or disconnect
sub process_line ($$) {
my ($self, $l) = @_;
+
+ # TODO: IMAP allows literals for big requests to upload messages
+ # (which we don't support) but maybe some big search queries use it.
+ # RFC 3501 9 (2) doesn't permit TAB or multiple SP
my ($tag, $req, @args) = parse_line('[ \t]+', 0, $l);
pop(@args) if (@args && !defined($args[-1]));
if (@args && uc($req) eq 'UID') {
$req .= "_".(shift @args);
}
my $res = eval {
- if (my $cmd = $self->can('cmd_'.lc($req // ''))) {
- defined($self->{-idle_tag}) ?
- "$self->{-idle_tag} BAD expected DONE\r\n" :
- $cmd->($self, $tag, @args);
- } elsif (uc($tag // '') eq 'DONE' && !defined($req)) {
- cmd_done($self, $tag);
+ if (defined(my $idle_tag = $self->{-idle_tag})) {
+ (uc($tag // '') eq 'DONE' && !defined($req)) ?
+ idle_done($self, $tag) :
+ "$idle_tag BAD expected DONE\r\n";
+ } elsif (my $cmd = $self->can('cmd_'.lc($req // ''))) {
+ if ($cmd == \&cmd_uid_search || $cmd == \&cmd_search) {
+ # preserve user-supplied quotes for search
+ (undef, @args) = split(/ search /i, $l, 2);
+ }
+ $cmd->($self, $tag, @args);
} else { # this is weird
auth_challenge_ok($self) //
($tag // '*') .
} elsif ($more) { # $self->{wbuf}:
$self->update_idle_time;
- # control passed to $more may be a GitAsyncCat object
+ # control passed to git_async_cat if $more == \undef
requeue_once($self) if !ref($more);
} else { # all done!
delete $self->{long_cb};
my $rbuf = $self->{rbuf} // \(my $x = '');
my $line = index($$rbuf, "\n");
while ($line < 0) {
- return $self->close if length($$rbuf) >= LINE_MAX;
- $self->do_read($rbuf, LINE_MAX, length($$rbuf)) or return;
+ if (length($$rbuf) >= LINE_MAX) {
+ $self->write(\"\* BAD request too long\r\n");
+ return $self->close;
+ }
+ $self->do_read($rbuf, LINE_MAX, length($$rbuf)) or
+ return uo2m_hibernate($self);
$line = index($$rbuf, "\n");
}
$line = substr($$rbuf, 0, $line + 1, '');
# for graceful shutdown in PublicInbox::Daemon:
sub busy {
my ($self, $now) = @_;
+ if (defined($self->{-idle_tag})) {
+ $self->write(\"* BYE server shutting down\r\n");
+ return; # not busy anymore
+ }
($self->{rbuf} || $self->{wbuf} || $self->not_idle_long($now));
}
sub close {
my ($self) = @_;
if (my $ibx = delete $self->{ibx}) {
- if (my $sock = $self->{sock}) {;
- $ibx->unsubscribe_unlock(fileno($sock));
- }
+ stop_idle($self, $ibx);
}
$self->SUPER::close; # PublicInbox::DS::close
}
# we're read-only, so SELECT and EXAMINE do the same thing
no warnings 'once';
*cmd_select = \&cmd_examine;
-*cmd_fetch = \&cmd_uid_fetch;
+
+package PublicInbox::IMAP_preauth;
+our @ISA = qw(PublicInbox::IMAP);
+
+sub logged_in { 0 }
1;