589ab02dc4
Ensure piece completion state is correct at startup
2016-02-17 17:20:21 +11:00
73208b97ca
Code comments
2016-02-17 17:09:43 +11:00
0f9f7ba01b
Add torrent.completedPieces bitmap
...
Reduce load on data.PieceComplete for torrents with lots of pieces, when reader position changes. Not sure of the improvement yet.
2016-02-17 00:00:55 +11:00
b514434581
Trigger piece request order changes on connections in a few spots
...
The client was requesting already obtained data in certain circumstances. This fixes it.
2016-02-11 17:57:57 +11:00
034c816934
torrent.piecePartiallyDownloaded incorrectly returned true when all the chunks were dirty
2016-02-10 00:46:54 +11:00
24b8b13d2d
Fix pending of already completed pieces when pending ranges of pieces
2016-02-08 21:36:50 +11:00
dab95b5431
Remove unreachable code
2016-02-08 03:01:54 +11:00
623d9e0e06
Rework requesting of shuffled pending pieces
2016-02-07 21:58:48 +11:00
dc215d083e
Export PieceStateChange through piece state change pubsub
2016-02-07 21:55:47 +11:00
0c28a0f1b1
Use new tracker package interface
2016-02-07 18:49:35 +11:00
77d6e9e5cb
Use new missinggo iterator style, and speed up torrent.connHasWantedPieces()
2016-02-07 01:22:31 +11:00
1b8bc1498c
Use missinggo.bitmap for tracking dirty chunks
2016-02-05 15:45:32 +11:00
6942f3e6d5
Changes to piece priority need to be published as a piece state change
...
This occurs because we're now able to deprioritize pieces after readers seek or close
2016-02-05 01:19:42 +11:00
63c73e18b3
Add File.Cancel
2016-02-05 01:18:54 +11:00
89f23c226c
Fix bug in readahead
2016-02-02 04:45:57 +11:00
430f26f726
Recalculate all piece priorities more efficiently
2016-02-01 22:06:13 +11:00
afa264e6c5
Reintroduce connection piece inclinations, and begin caching piece priorities
2016-02-01 21:11:41 +11:00
3ed628356b
Use missinggo.Event for connection closing event
2016-02-01 21:08:52 +11:00
b90dacd324
Abstract the pendingPieces as a bitmap
2016-02-01 01:46:28 +11:00
ae9bba2611
Fix downloading of unwanted chunks, and write out downloaded chunks using the connection loop goroutine
...
Doing the writes with the connection loop goroutine prevents memory use blowing out when chunks aren't written out quickly enough.
2016-01-28 05:54:48 +11:00
90348f6a48
Add pending write helpers
2016-01-25 07:22:33 +11:00
94d764e6be
Readers changing wasn't triggering waitAccept
...
Caused occasional stall in TestSeedAfterDownloading.
2016-01-25 04:14:19 +11:00
0ad4dda9fc
Move fillRequests and replenishConnRequests into connection.go
2016-01-24 15:21:17 +11:00
2f40c48d37
Fix the download/prioritize piece functions
...
This involves adding a pendingPieces field to torrent.
2016-01-19 01:28:56 +11:00
06445f2a1e
It's working and the tests are usually passing
...
I still need to handle "prefetch"-style downloading, and some functions haven't been committed to force this issue.
2016-01-18 18:35:14 +11:00
89a6a20af6
Finally make Torrent.torrent private
...
A lot of code was using unintentionally exported stuff from the embedded *torrent in Torrent.
2016-01-17 00:14:15 +11:00
2beb5f8bd4
Track dirty chunks, instead of pending chunk specs
...
This is the first step toward having purely Reader-based priorities. If a chunk is pending, that currently implies that we want to download it. I want to move that kind of state out to the readers.
2016-01-13 17:11:59 +11:00
153c13db43
Revert "Merge pull request #54 from zhulik/master"
...
This reverts commit 5cf75b869c09050ff2c5
2016-01-06 12:19:49 +11:00
8637086ef9
Merge branch 'master' of https://github.com/anacrolix/torrent
2016-01-04 15:39:08 +03:00
8346ff81f7
Switch to int from peer_protocol.Integer for piece index in a lot of places
2016-01-04 22:34:24 +11:00
be8e44da3c
peersKey is a public type now
2015-12-27 15:27:32 +03:00
86c4c609f1
Expose Torrent.SetDisplayName
2015-12-12 14:03:04 +11:00
79eb1ca0d5
torrent.DisplayName should not be exported
2015-11-22 18:44:08 +11:00
d01c6221e6
Remove sync.Cond from piece
...
It's not trivial to ensure that anyone waiting on the piece Cond will wake on events like the torrent being dropped from the client.
2015-11-06 00:40:16 +11:00
c7cb1009a5
Fix #33 .
2015-10-20 17:53:12 +11:00
b3a8020401
Store pieces inplace in torrent.Pieces
...
This should save some allocation overhead, especially for torrents that have 20k+ pieces.
2015-10-16 22:10:03 +11:00
b0b5794890
Create a pieceStore interface, and merge in my httpfile backend, and replace data/blob
...
data/blob was aging, and had severe performance problems. It's now possible to use missinggo/filecache as a data backend to pieceStore which is better tested and performs excellently.
2015-10-04 00:22:46 +10:00
b7a8bb7570
Simplify the torrent Data interface
...
None of the methods are optional anymore. Removed the nasty wrappers for some data implementations. Moved data.Data back into the main torrent package. Should make it much easier to understand for people implementing their own Data implementations.
2015-10-02 00:09:04 +10:00
58c997210e
Track peer sources with a dict, and don't exceed high water mark
2015-09-28 15:30:13 +10:00
53259397fc
Initialize connection piece priorities lazily, and using a pool
2015-09-26 17:27:35 +10:00
52b0705956
Include num pieces in torrent status
2015-09-25 22:05:32 +10:00
f9ed9c4c04
Avoid frequent allocation of piece request order iterator
2015-09-20 11:29:52 +10:00
eebd09c0fe
Allow subscribing to torrent piece state changes
2015-09-06 12:33:22 +10:00
ef098c47ac
Needlessly convoluted
2015-08-19 13:29:01 +10:00
085be622a0
Replacing pendingWrites WaitGroup with Mutex/Cond/int
...
Apparently I've been using WaitGroups wrong all along.
2015-08-05 02:40:46 +10:00
85a0fe6b61
Drop peers for sending too many unwanted pieces
...
So the algorithm is now, drop the worst half if they get too old, and drop any connection if it's ratio is poor.
2015-08-04 01:32:45 +10:00
d077fed72d
Some utils moved to missinggo
2015-08-04 00:29:01 +10:00
0600c3b5e3
Several speedups in logic
2015-07-17 21:07:01 +10:00
7e9fe4f447
Store chunk data without holding client lock
2015-07-15 16:00:59 +10:00
c018c660f0
Allow chunk size to be specified per torrent
2015-07-15 15:31:18 +10:00
Matt Joiner