X-Git-Url: https://git.cworth.org/git?a=blobdiff_plain;f=lib%2Fsup%2Fimap.rb;h=4eb13f4ed619d3259df4894eec4a8b66c11c326e;hb=6461b6b5c80b216627127837942b761a151317b5;hp=c1ba2bf6c4b366d1d60c1d79bcf22d6a5d81163d;hpb=df62e1a6e52f6b5c0dd542e79307aa7cdcc0a88d;p=sup diff --git a/lib/sup/imap.rb b/lib/sup/imap.rb index c1ba2bf..4eb13f4 100644 --- a/lib/sup/imap.rb +++ b/lib/sup/imap.rb @@ -3,40 +3,44 @@ require 'net/imap' require 'stringio' require 'time' require 'rmail' +require 'cgi' -## fucking imap fucking sucks. what the FUCK kind of committee of -## dunces designed this shit. +## TODO: remove synchronized method protector calls; use a Monitor instead +## (ruby's reentrant mutex) + +## fucking imap fucking sucks. what the FUCK kind of committee of dunces +## designed this shit. ## ## imap talks about 'unique ids' for messages, to be used for -## cross-session identification. great---just what sup needs! except -## it turns out the uids can be invalidated every time the -## 'uidvalidity' value changes on the server, and 'uidvalidity' can -## change without restriction. it can change any time you log in. it -## can change EVERY time you log in. of course the imap spec "strongly -## recommends" that it never change, but there's nothing to stop -## people from just setting it to the current timestamp, and in fact -## that's exactly what the one imap server i have at my disposal -## does. thus the so-called uids are absolutely useless and imap -## provides no cross-session way of uniquely identifying a -## message. but thanks for the "strong recommendation", guys! +## cross-session identification. great---just what sup needs! except it +## turns out the uids can be invalidated every time the 'uidvalidity' +## value changes on the server, and 'uidvalidity' can change without +## restriction. it can change any time you log in. it can change EVERY +## time you log in. of course the imap spec "strongly recommends" that it +## never change, but there's nothing to stop people from just setting it +## to the current timestamp, and in fact that's EXACTLY what the one imap +## server i have at my disposal does. thus the so-called uids are +## absolutely useless and imap provides no cross-session way of uniquely +## identifying a message. but thanks for the "strong recommendation", +## guys! ## ## so right now i'm using the 'internal date' and the size of each ## message to uniquely identify it, and i scan over the entire mailbox ## each time i open it to map those things to message ids. that can be -## slow for large mailboxes, and we'll just have to hope that there -## are no collisions. ho ho! a perfectly reasonable solution! +## slow for large mailboxes, and we'll just have to hope that there are +## no collisions. ho ho! a perfectly reasonable solution! ## ## and here's another thing. check out RFC2060 2.2.2 paragraph 5: ## -## A client MUST be prepared to accept any server response at all times. -## This includes server data that was not requested. +## A client MUST be prepared to accept any server response at all +## times. This includes server data that was not requested. ## -## yeah. that totally makes a lot of sense. and once again, the idiocy -## of the spec actually happens in practice. you'll request flags for -## one message, and get it interspersed with a random bunch of flags -## for some other messages, including a different set of flags for the -## same message! totally ok by the imap spec. totally retarded by any -## other metric. +## yeah. that totally makes a lot of sense. and once again, the idiocy of +## the spec actually happens in practice. you'll request flags for one +## message, and get it interspersed with a random bunch of flags for some +## other messages, including a different set of flags for the same +## message! totally ok by the imap spec. totally retarded by any other +## metric. ## ## fuck you, imap committee. you managed to design something nearly as ## shitty as mbox but goddamn THIRTY YEARS LATER. @@ -62,39 +66,28 @@ class IMAP < Source @username = username @password = password @imap = nil - @imap_ids = {} + @imap_state = {} @ids = [] @last_scan = nil - @labels = (labels || []).freeze + @labels = ((labels || []) - LabelManager::RESERVED_LABELS).uniq.freeze @say_id = nil @mutex = Mutex.new end def self.suggest_labels_for path - if path =~ /inbox/i - [path.intern] - else - [] - end + path =~ /([^\/]*inbox[^\/]*)/i ? [$1.downcase.intern] : [] end def host; @parsed_uri.host; end def port; @parsed_uri.port || (ssl? ? 993 : 143); end def mailbox x = @parsed_uri.path[1..-1] - x.nil? || x.empty? ? 'INBOX' : x + (x.nil? || x.empty?) ? 'INBOX' : CGI.unescape(x) end def ssl?; @parsed_uri.scheme == 'imaps' end - def check - ids = - @mutex.synchronize do - unsynchronized_scan_mailbox - @ids - end - - start = ids.index(cur_offset || start_offset) or raise OutOfSyncSourceError, "Unknown message id #{cur_offset || start_offset}." - end + def check; end # do nothing because anything we do will be too slow, + # and we'll catch the errors later. ## is this necessary? TODO: remove maybe def == o; o.is_a?(IMAP) && o.uri == self.uri && o.username == self.username; end @@ -104,23 +97,41 @@ class IMAP < Source end def load_message id - RMail::Parser.read raw_full_message(id) + RMail::Parser.read raw_message(id) + end + + def each_raw_message_line id + StringIO.new(raw_message(id)).each { |l| yield l } end def raw_header id unsynchronized_scan_mailbox - header, flags = get_imap_fields id, 'RFC822.HEADER', 'FLAGS' - ## very bad. this is very very bad. very bad bad bad. - header = header + "Status: RO\n" if flags.include? :Seen # fake an mbox-style read header # TODO: improve source-marked-as-read reporting system + header, flags = get_imap_fields id, 'RFC822.HEADER' header.gsub(/\r\n/, "\n") end synchronized :raw_header - def raw_full_message id + def raw_message id unsynchronized_scan_mailbox get_imap_fields(id, 'RFC822').first.gsub(/\r\n/, "\n") end - synchronized :raw_full_message + synchronized :raw_message + + def mark_as_deleted ids + ids = [ids].flatten # accept single arguments + unsynchronized_scan_mailbox + imap_ids = ids.map { |i| @imap_state[i] && @imap_state[i][:id] }.compact + return if imap_ids.empty? + @imap.store imap_ids, "+FLAGS", [:Deleted] + end + synchronized :mark_as_deleted + + def expunge + @imap.expunge + unsynchronized_scan_mailbox true + true + end + synchronized :expunge def connect return if @imap @@ -128,27 +139,31 @@ class IMAP < Source end synchronized :connect - def scan_mailbox - return if @last_scan && (Time.now - @last_scan) < SCAN_INTERVAL + def scan_mailbox force=false + return if !force && @last_scan && (Time.now - @last_scan) < SCAN_INTERVAL last_id = safely do @imap.examine mailbox @imap.responses["EXISTS"].last end @last_scan = Time.now + @ids = [] if force return if last_id == @ids.length range = (@ids.length + 1) .. last_id Redwood::log "fetching IMAP headers #{range}" - fetch(range, ['RFC822.SIZE', 'INTERNALDATE']).each do |v| + fetch(range, ['RFC822.SIZE', 'INTERNALDATE', 'FLAGS']).each do |v| id = make_id v @ids << id - @imap_ids[id] = v.seqno + @imap_state[id] = { :id => v.seqno, :flags => v.attr["FLAGS"] } end + Redwood::log "done fetching IMAP headers" end synchronized :scan_mailbox def each + return unless start_offset + ids = @mutex.synchronize do unsynchronized_scan_mailbox @@ -157,10 +172,19 @@ class IMAP < Source start = ids.index(cur_offset || start_offset) or raise OutOfSyncSourceError, "Unknown message id #{cur_offset || start_offset}." - start.upto(ids.length - 1) do |i| + start.upto(ids.length - 1) do |i| id = ids[i] - self.cur_offset = id - yield id, @labels + state = @mutex.synchronize { @imap_state[id] } or next + self.cur_offset = id + labels = { :Flagged => :starred, + :Deleted => :deleted + }.inject(@labels) do |cur, (imap, sup)| + cur + (state[:flags].include?(imap) ? [sup] : []) + end + + labels += [:unread] unless state[:flags].include?(:Seen) + + yield id, labels end end @@ -172,7 +196,7 @@ class IMAP < Source def end_offset unsynchronized_scan_mailbox - @ids.last + @ids.last + 1 end synchronized :end_offset @@ -201,11 +225,11 @@ private def unsafe_connect say "Connecting to IMAP server #{host}:#{port}..." - ## apparently imap.rb does a lot of threaded stuff internally and - ## if an exception occurs, it will catch it and re-raise it on the - ## calling thread. but i can't seem to catch that exception, so - ## i've resorted to initializing it in its own thread. surely - ## there's a better way. + ## apparently imap.rb does a lot of threaded stuff internally and if + ## an exception occurs, it will catch it and re-raise it on the + ## calling thread. but i can't seem to catch that exception, so i've + ## resorted to initializing it in its own thread. surely there's a + ## better way. exception = nil ::Thread.new do begin @@ -213,14 +237,16 @@ private @imap = Net::IMAP.new host, port, ssl? say "Logging in..." - ## although RFC1730 claims that "If an AUTHENTICATE command - ## fails with a NO response, the client may try another", in - ## practice it seems like they can also send a BAD response. + ## although RFC1730 claims that "If an AUTHENTICATE command fails + ## with a NO response, the client may try another", in practice + ## it seems like they can also send a BAD response. begin + raise Net::IMAP::NoResponseError unless @imap.capability().member? "AUTH=CRAM-MD5" @imap.authenticate 'CRAM-MD5', @username, @password rescue Net::IMAP::BadResponseError, Net::IMAP::NoResponseError => e Redwood::log "CRAM-MD5 authentication failed: #{e.class}. Trying LOGIN auth..." begin + raise Net::IMAP::NoResponseError unless @imap.capability().member? "AUTH=LOGIN" @imap.authenticate 'LOGIN', @username, @password rescue Net::IMAP::BadResponseError, Net::IMAP::NoResponseError => e Redwood::log "LOGIN authentication failed: #{e.class}. Trying plain-text LOGIN..." @@ -253,18 +279,34 @@ private %w(RFC822.SIZE INTERNALDATE).each do |w| raise FatalSourceError, "requested data not in IMAP response: #{w}" unless imap_stuff.attr[w] end - + msize, mdate = imap_stuff.attr['RFC822.SIZE'] % 10000000, Time.parse(imap_stuff.attr["INTERNALDATE"]) sprintf("%d%07d", mdate.to_i, msize).to_i end def get_imap_fields id, *fields - imap_id = @imap_ids[id] or raise OutOfSyncSourceError, "Unknown message id #{id}" + raise OutOfSyncSourceError, "Unknown message id #{id}" unless @imap_state[id] - retried = false + imap_id = @imap_state[id][:id] result = fetch(imap_id, (fields + ['RFC822.SIZE', 'INTERNALDATE']).uniq).first got_id = make_id result - raise OutOfSyncSourceError, "IMAP message mismatch: requested #{id}, got #{got_id}." unless got_id == id + + ## I've turned off the following sanity check because Microsoft + ## Exchange fails it. Exchange actually reports two different + ## INTERNALDATEs for the exact same message when queried at different + ## points in time. + ## + ## RFC2060 defines the semantics of INTERNALDATE for messages that + ## arrive via SMTP for via various IMAP commands, but states that + ## "All other cases are implementation defined.". Great, thanks guys, + ## yet another useless field. + ## + ## Of course no OTHER imap server I've encountered returns DIFFERENT + ## values for the SAME message. But it's Microsoft; what do you + ## expect? If their programmers were any good they'd be working at + ## Google. + + # raise OutOfSyncSourceError, "IMAP message mismatch: requested #{id}, got #{got_id}." unless got_id == id fields.map { |f| result.attr[f] or raise FatalSourceError, "empty response from IMAP server: #{f}" } end