acbd61e93d
optimization. should *greatly* reduce cpu usage. #136
11 years ago
fa386c250d
recheck piece just before sending to other torrent peer. fix #15 .
11 years ago
a9f53cce98
trying to reduce the occurrences of "INCOMING ERROR: height > getBestHeight"
11 years ago
08cadbb78e
enough of stack smashing segfault. openssl is required, period. fix #105 . fix #97 . fix #27 .
11 years ago
e3f7dfe811
apply patch from http://sourceforge.net/p/cygwin-ports/libtorrent-rasterbar/ci/master/tree/0.16.7-cygwin.patch
11 years ago
85be748663
trying to improve torrent connectivity in the long run: relax peer banning.
...
don't preserve banned_list on startup. peers could have been banned due to changes in
block chain (invalid key or height). we shouldn't keep them banned forever.
somewhat related to #15 .
11 years ago
35fa3832e8
bugfix. this was a tough one. because the observer was already dequeued
...
from m_transactions, we need to properly terminate it in some way. usually
we go through the end of the function, where reply() does done(). however
there are some error cases where function may returns sooner. this might be
unlikely to occur in original upstream code (missing 'r' key, missing 'id'),
but with IP enforcing it is not so uncommon. hopefully this will fix #32 .
11 years ago
4ec3817120
printf call expects long int
11 years ago
25128d93dd
refresh mention resource in dht
11 years ago
3e24404d7e
Fixed warning for snprintf call
11 years ago
5d11289b4c
fix some 'format not a string literal's
11 years ago
8e80324c36
try to reduce the level of useless dht tracker requests by not announcing empty torrents we do not follow.
11 years ago
8287b2b1e1
use a single database for all swarms of the same libtorrent session.
...
upgrade resume file to version 2 and enforce on loading (to force existing clients to download all pieces again).
fixes #24 .
11 years ago
60db413630
less noisy
11 years ago
2dd4c62b0e
increase active downloads. tests needed.
11 years ago
1ae5745151
another we_dont_have() due to disk read error.
...
the idea is to help recovering faster from db corruption problems.
11 years ago
8328c7222f
use we_dont_have to recover from corrupt swarm database.
...
eventually we will download the pieces again (it works)
11 years ago
ca1eb9f3c2
update "interesting" upon receiving bitfield
11 years ago
8dbae988ec
dont expire dht items if block chain is invalid
11 years ago
bcc7ee4747
increase dht upload rate.
...
explains some avatars not being served on timeline updates. client really needs to start caching this.
11 years ago
aa422d9b4e
increase number of active torrents
11 years ago
9aec1f6c73
libtorrent fix from upstream
11 years ago
77cfe2cd6c
libtorrent fix: should prevent the infinite sequence of auto manager triggering
11 years ago
80f389239b
too strict assert causes twisterd to abort when network goes down
11 years ago
c3bea0a3dd
If bitcoin network goes down, pause libtorrent to prevent pieces being falsely rejected because height > bestHeight.
...
Also, if bitcoin network is down, try to add nodes from DHT bucket tables back to the bitcoin network.
11 years ago
54dc49c191
enable expiration of dht items
11 years ago
a654c48153
one more assert to complement previous commit
11 years ago
d8b8ff4b4f
trying to fix/understand assertion failure
...
version: 1.0.0.0
$Rev: 7680 $
file: 'peer_connection.cpp'
line: 5715
function: void libtorrent::peer_connection::check_invariant() const
expression: m_outstanding_bytes >= 0
11 years ago
2c256b6edb
implement piece verification using old signatures. now i can reenable banning peers due to bad pieces.
11 years ago
8ec4b3e16e
i can't get this conversion from time_duration to string right. i give up.
11 years ago
b7fb1f3d3c
strange android/boost compilation fix
11 years ago
658ed6054a
export number of known torrent peers (list_peers) to the dht tracker resource and use it
...
as estimative to the number of followers. at least we get the number of *online* followers.
11 years ago
cd4147d772
increase refresh_storage sleep when zero refreshable items
11 years ago
a4695ba9e1
delay libtorrent initialization until we have valid blocks
11 years ago
80f23d56e8
change "values" to "data" in getData to differentiate from tracker values (which are not authenticated/signed)
11 years ago
5a9613297a
dht storage refresh is spread over time to avoid bursts of putData
11 years ago
549ba353dd
yet another ThreadMaintainDHTNodes heuristic improvement
11 years ago
2cd0b0a881
rescandirectmsgs: rescan all streams of users we follow for new and old directmessages
11 years ago
c9e2b88920
try another improvement to local dht tracker updating: currently connected nodes or recently connected (4 hours)
11 years ago
624b5f8b49
improved debugging of get/put dht keys
11 years ago
d363d2f788
remove unneeded optimization: m_data_callback is called multiple times (for every reply) so asking ourselves doesn't hurt.
...
this might fix the assertion which is failing on observer destructor.
11 years ago
40803b910c
improve last commit: only announce to local dht tracker if we ever connected to peer.
11 years ago
ac76c967ae
do not add peers with high failcount to the local dht tracker
11 years ago
c0ec1f4afe
Until old public key is properly used, disable banning torrent peers due to bad piece hashes.
...
+note: torrent.cpp line 3286 (function piece_failed), iteration to ban peers is disabled (continue)
11 years ago
0150dd2f7b
support to disable loading/refreshing of expired dht keys
11 years ago
a19c4f231f
return false when packet not sent to ourselves
11 years ago
15d7fd07b0
android compiling
11 years ago
7e1726ba8c
implement expiring dht items (but it is not enabled yet)
11 years ago
b7aac66551
android fixes
11 years ago
52a3ecda88
fix compilation error (m_sent_bitfield) without libtorrent debug
11 years ago
Miguel Freitas