require 'net/imap'
require 'stringio'
require 'time'
+require 'rmail'
## 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
## 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!
-
-## fuck you, imap committee. you managed to design something as shitty
-## as mbox but goddamn THIRTY YEARS LATER.
-
+##
+## 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.
+##
+## 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.
module Redwood
class IMAP < Source
SCAN_INTERVAL = 60 # seconds
- attr_reader_cloned :labels
+ ## upon these errors we'll try to rereconnect a few times
+ RECOVERABLE_ERRORS = [ Errno::EPIPE, Errno::ETIMEDOUT, OpenSSL::SSL::SSLError ]
+
attr_accessor :username, :password
+ yaml_properties :uri, :username, :password, :cur_offset, :usual,
+ :archived, :id, :labels
- def initialize uri, username, password, last_idate=nil, usual=true, archived=false, id=nil
+ def initialize uri, username, password, last_idate=nil, usual=true, archived=false, id=nil, labels=[]
raise ArgumentError, "username and password must be specified" unless username && password
raise ArgumentError, "not an imap uri" unless uri =~ %r!imaps?://!
@imap_ids = {}
@ids = []
@last_scan = nil
- @labels = [:unread]
- @labels << :inbox unless archived?
- @labels << mailbox.intern unless mailbox =~ /inbox/i
+ @labels = (labels || []).freeze
+ @say_id = nil
@mutex = Mutex.new
end
+ def self.suggest_labels_for path
+ if path =~ /inbox/i
+ [path.intern]
+ else
+ []
+ end
+ end
+
def host; @parsed_uri.host; end
def port; @parsed_uri.port || (ssl? ? 993 : 143); end
def mailbox
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
+
+ ## is this necessary? TODO: remove maybe
+ def == o; o.is_a?(IMAP) && o.uri == self.uri && o.username == self.username; end
+
def load_header id
MBox::read_header StringIO.new(raw_header(id))
end
end
def raw_header id
- @mutex.synchronize do
- connect
- header, flags = get_imap_fields id, 'RFC822.HEADER', 'FLAGS'
- header = "Status: RO\n" + header if flags.include? :Seen # fake an mbox-style read header
- header.gsub(/\r\n/, "\n")
- end
+ 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.gsub(/\r\n/, "\n")
end
+ synchronized :raw_header
def raw_full_message id
- @mutex.synchronize do
- connect
- get_imap_fields(id, 'RFC822').first.gsub(/\r\n/, "\n")
- end
+ unsynchronized_scan_mailbox
+ get_imap_fields(id, 'RFC822').first.gsub(/\r\n/, "\n")
end
+ synchronized :raw_full_message
def connect
- return false if broken?
- return true if @imap
+ return if @imap
+ safely { } # do nothing!
+ end
+ synchronized :connect
- say "Connecting to IMAP server #{host}:#{port}..."
+ def scan_mailbox
+ return if @last_scan && (Time.now - @last_scan) < SCAN_INTERVAL
+ last_id = safely do
+ @imap.examine mailbox
+ @imap.responses["EXISTS"].last
+ end
+ @last_scan = Time.now
+
+ 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|
+ id = make_id v
+ @ids << id
+ @imap_ids[id] = v.seqno
+ end
+ end
+ synchronized :scan_mailbox
+
+ def each
+ 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}."
- ## ok, this is FUCKING ANNOYING.
- ##
- ## what imap.rb likes to do is, if an exception occurs, catch it
- ## and re-raise it on the calling thread. seems reasonable. but
- ## what that REALLY means is that the only way to reasonably
- ## initialize imap is in its own thread, because otherwise, you
- ## will never be able to catch the exception it raises on the
- ## calling thread, and the backtrace will not make any sense at
- ## all, and you will waste HOURS of your life on this fucking
- ## problem.
- ##
- ## FUCK!!!!!!!!!
+ start.upto(ids.length - 1) do |i|
+ id = ids[i]
+ self.cur_offset = id
+ yield id, @labels
+ end
+ end
+
+ def start_offset
+ unsynchronized_scan_mailbox
+ @ids.first
+ end
+ synchronized :start_offset
+ def end_offset
+ unsynchronized_scan_mailbox
+ @ids.last
+ end
+ synchronized :end_offset
+
+ def pct_done; 100.0 * (@ids.index(cur_offset) || 0).to_f / (@ids.length - 1).to_f; end
+
+private
+
+ def fetch ids, fields
+ results = safely { @imap.fetch ids, fields }
+ good_results =
+ if ids.respond_to? :member?
+ results.find_all { |r| ids.member?(r.seqno) && fields.all? { |f| r.attr.member?(f) } }
+ else
+ results.find_all { |r| ids == r.seqno && fields.all? { |f| r.attr.member?(f) } }
+ end
+
+ if good_results.empty?
+ raise FatalSourceError, "no IMAP response for #{ids} containing all fields #{fields.join(', ')} (got #{results.size} results)"
+ elsif good_results.size < results.size
+ Redwood::log "Your IMAP server sucks. It sent #{results.size} results for a request for #{good_results.size} messages. What are you using, Binc?"
+ end
+
+ good_results
+ end
+
+ 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.
exception = nil
- Redwood::reporting_thread do
+ ::Thread.new do
begin
#raise Net::IMAP::ByeResponseError, "simulated imap failure"
@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.
begin
@imap.authenticate 'CRAM-MD5', @username, @password
rescue Net::IMAP::BadResponseError, Net::IMAP::NoResponseError => e
- say "CRAM-MD5 authentication failed: #{e.class}"
+ Redwood::log "CRAM-MD5 authentication failed: #{e.class}. Trying LOGIN auth..."
begin
@imap.authenticate 'LOGIN', @username, @password
rescue Net::IMAP::BadResponseError, Net::IMAP::NoResponseError => e
- say "LOGIN authentication failed: #{e.class}"
+ Redwood::log "LOGIN authentication failed: #{e.class}. Trying plain-text LOGIN..."
@imap.login @username, @password
end
end
- scan_mailbox
say "Successfully connected to #{@parsed_uri}."
- rescue SocketError, Net::IMAP::Error, SourceError => e
+ rescue Exception => e
exception = e
ensure
shutup
end
end.join
- die_from exception, :while => "connecting" if exception
- end
-
- def each
- @mutex.synchronize { connect or raise SourceError, broken_msg }
-
- start = @ids.index(cur_offset || start_offset) or die_from "Unknown message id #{cur_offset || start_offset}.", :suggest_rebuild => true # couldn't find the most recent email
-
- start.upto(@ids.length - 1) do |i|
- id = @ids[i]
- self.cur_offset = id
- yield id, labels
- end
+ raise exception if exception
end
- def start_offset
- @mutex.synchronize { connect }
- @ids.first
- end
-
- def end_offset
- @mutex.synchronize do
- begin
- connect
- scan_mailbox
- rescue SocketError, Net::IMAP::Error => e
- die_from e, :while => "scanning mailbox"
- end
- end
- @ids.last
- end
-
- def pct_done; 100.0 * (@ids.index(cur_offset) || 0).to_f / (@ids.length - 1).to_f; end
-
-private
-
def say s
@say_id = BufferManager.say s, @say_id if BufferManager.instantiated?
Redwood::log s
@say_id = nil
end
- def scan_mailbox
- return if @last_scan && (Time.now - @last_scan) < SCAN_INTERVAL
-
- @imap.examine mailbox
- last_id = @imap.responses["EXISTS"].last
- @last_scan = Time.now
- return if last_id == @ids.length
- Redwood::log "fetching IMAP headers #{(@ids.length + 1) .. last_id}"
- values = @imap.fetch((@ids.length + 1) .. last_id, ['RFC822.SIZE', 'INTERNALDATE'])
- values.each do |v|
- id = make_id v
- @ids << id
- @imap_ids[id] = v.seqno
- end
- end
-
- def die_from e, opts={}
- @imap = nil
-
- message =
- case e
- when Exception
- "Error while #{opts[:while]}: #{e.message.chomp} (#{e.class.name})."
- when String
- e
- end
-
- message += " It is likely that messages have been deleted from this IMAP mailbox. Please run sup-import --rebuild #{to_s} to correct this problem." if opts[:suggest_rebuild]
-
- self.broken_msg = message
- Redwood::log message
- BufferManager.flash "Error communicating with IMAP server. See log for details." if BufferManager.instantiated?
- raise SourceError, message
- end
-
- ## build a fake unique id
def make_id imap_stuff
# use 7 digits for the size. why 7? seems nice.
+ %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}"
+
+ retried = false
+ 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
+
+ fields.map { |f| result.attr[f] or raise FatalSourceError, "empty response from IMAP server: #{f}" }
+ end
+
+ ## execute a block, connected if unconnected, re-connected up to 3
+ ## times if a recoverable error occurs, and properly dying if an
+ ## unrecoverable error occurs.
+ def safely
retries = 0
- f = nil
- imap_id = @imap_ids[id] or die_from "Unknown message id #{id}.", :suggest_rebuild => true
begin
- f = @imap.fetch imap_id, (fields + ['RFC822.SIZE', 'INTERNALDATE']).uniq
- got_id = make_id f[0]
- die_from "IMAP message mismatch: requested #{id}, got #{got_id}.", :suggest_rebuild => true unless id == got_id
- rescue SocketError, Net::IMAP::Error => e
- die_from e, :while => "communicating with IMAP server"
- rescue Errno::EPIPE
- if (retries += 1) <= 3
- @imap = nil
- connect
- retry
+ begin
+ unsafe_connect unless @imap
+ yield
+ rescue *RECOVERABLE_ERRORS => e
+ if (retries += 1) <= 3
+ @imap = nil
+ Redwood::log "got #{e.class.name}: #{e.message.inspect}"
+ sleep 2
+ retry
+ end
+ raise
end
+ rescue SocketError, Net::IMAP::Error, SystemCallError, IOError, OpenSSL::SSL::SSLError => e
+ raise FatalSourceError, "While communicating with IMAP server (type #{e.class.name}): #{e.message.inspect}"
end
- die_from "Null IMAP field '#{field}' for message with id #{id} imap id #{imap_id}." if f.nil?
-
- fields.map { |field| f[0].attr[field] }
end
-end
-Redwood::register_yaml(IMAP, %w(uri username password cur_offset usual archived id))
+end
end