ogb912/opensim
1
0
Fork 0
mirror of https://github.com/opensim/opensim.git synced 2026-05-21 07:45:41 +08:00
Code Issues Packages Projects Releases Wiki Activity
25,254 Commits 5 Branches 93 Tags
242bb425f37ebe258d2df46babc4e4553a4ad33e
Commit Graph

2699 Commits

Author SHA1 Message Date
Melanie Thielker
8014ce7f1a Fix a nullref in a debug message. If we have no SceneAgent, we can't output data from it. 2014-09-15 23:15:27 +02:00
Melanie Thielker
d43d3df724 Reinstate closing the client if CloseAgen returns false. This was part 
of the last Ubit patch.
 2014-09-14 20:46:22 +02:00
UbitUmarov
3d9f25a19d revert changes to m_pendingCache and remove client close if there is no 
SP, as requested
 2014-09-14 02:28:42 +01:00
UbitUmarov
908e0e1cef do not use a blind expire of m_pendingCache. Entries must be valid while 
respective HandleUseCircuitCode, no matter how long it takes, so a viewer
 retry doesnt start a overlapping one. It HandleUseCircuitCode
 responsability to always remove entries, even on fails. Remove a client
 even if m_scene.CloseAgent fails.
 2014-09-13 20:37:20 +01:00
Melanie Thielker
42b2248960 If a client's SceneAgent is null, that client is not fully logged in or 
crashed while logging out. Don't reuse it, create a new one.
 2014-09-13 00:49:32 +02:00
UbitUmarov
0514679b15 reduce MIN_CALLBACK_MS guard time. A value too high introduces a extra 
throttle and makes more packets to be sent in bursts and not in steady
 state flow.
 2014-09-02 18:51:49 +01:00
UbitUmarov
6611e75a2b void time change between gettime calls 2014-09-02 16:11:46 +01:00
UbitUmarov
50433e089b *needs testing, not that good* change throttles math using floats and not 
int64, etc. Limite brust bytes to the total rate client requested times a
look ahead estimation time, Avoid queues starvation with updates waiting...
 2014-09-02 15:48:59 +01:00
UbitUmarov
18de5c8a2d make use of it in GetMeshModule, and meshs are Assets not Task itens, 
reduce MeshModule impact on udp rate
 2014-08-29 16:30:49 +01:00
UbitUmarov
7351d92a76 add method to get a category throttle rate 2014-08-29 16:19:30 +01:00
UbitUmarov
fd0a635302 reduce burst 2014-08-29 15:13:46 +01:00
UbitUmarov
438798202f replace the tick() by a limit on the maximum number of tokens that can be 
acumulated ( variable named BurtRate, not exactly a rate...)
 2014-08-29 13:26:30 +01:00
UbitUmarov
6f590c2f79 disable first drip 2014-08-29 00:05:57 +01:00
UbitUmarov
986863a0cd try to reduce insane high data rate udp bursts. This needs testing on a 
region with a lot of contents. Should not affect much average rates.
 2014-08-28 22:26:03 +01:00
UbitUmarov
894b5c10c6 remove misplaced comment 2014-08-28 18:54:08 +01:00
UbitUmarov
9839904ebe try to make sense of throttle rate limits 2014-08-28 18:51:55 +01:00
UbitUmarov
5a2d4fd47f add some functions for estimation of number of bytes that can be send in a category in specified time 2014-08-28 16:49:32 +01:00
UbitUmarov
ff4df688ba send all presences terseupdates in same batch 2014-08-28 12:55:31 +01:00
UbitUmarov
ade4bf69b1 if we send wearables with ThrottleOutPacketType.HighPriority, then we 
should send other avatarinformation with same priority on same Task category ( plus cleanup )
 2014-08-28 10:38:31 +01:00
UbitUmarov
1e888d61ca enqueue also if m_nextPackets[category] is not null. This is really the 
top element of a category queue, equivalente to using a queue.peek() if
 avaiable
 2014-08-27 01:33:09 +01:00
UbitUmarov
92b0b27caf dont append acks to a resend packet 2014-08-26 17:42:44 +01:00
UbitUmarov
6857afe2a4 *test* back to no sog kills, little retouch on attachment updates, 
NameValue AttachItemID is a root part thing only
 2014-08-25 23:31:04 +01:00
UbitUmarov
75cbe620c0 *test* cleanup the bugs nest 2014-08-25 22:45:31 +01:00
UbitUmarov
c24601bc6a change enconding of attachment updates NameValue and State fields 2014-08-25 21:49:39 +01:00
UbitUmarov
f4a6be8654 clean sendKill a bit, remove a IsChild that is set too early in a calling 
path
 2014-08-24 00:44:14 +01:00
UbitUmarov
d55041d934 on updates, send rotations using livomv Quaternion.toBytes() again 2014-08-23 21:13:30 +01:00
UbitUmarov
63d1916f51 Remove entities from updates queues on kill. Do it sync so enqueues after 
the kill work
 2014-08-23 20:59:11 +01:00
UbitUmarov
5bf145a397 add a direct sendpartfullUpdate to send a full object update to a part, 
optionally overriding its parentID. check what it does to attachments
 2014-08-21 00:49:10 +01:00
UbitUmarov
ea1c232f92 revert droping udp packet resends after 6 retries, keep resending. 2014-08-21 00:08:43 +01:00
UbitUmarov
542118adf1 remove from use the UpdatesResend on resending udp packets. Just resend 
the UDP packet. Also just loose packets we tried to send 6 times already
(ll says 3) A viewer may just beeing ignoring them, or then the link is just dead.
 2014-08-20 18:55:06 +01:00
UbitUmarov
919aef1573 send zero velocity again on avatar full update or its ugly 2014-08-19 10:03:04 +01:00
UbitUmarov
8f0d35e59a fix the encoding of rotation in updates, not just using the next field to 
override w bytes. ( specially having it commented )
 2014-08-19 09:04:20 +01:00
UbitUmarov
505cbf9983 still another ... 2014-08-19 03:58:16 +01:00
UbitUmarov
4c46ebdbf5 fix a missed blocking of sending updates the the new attach points above 
hud indexes
 2014-08-19 03:47:26 +01:00
UbitUmarov
1edaf29149 NextAnimationSequenceNumber be a udpserver variable with random start 2014-08-16 13:43:26 +01:00
UbitUmarov
0ae8fed4c2 reduce ping cliping lower limit 2014-08-14 21:18:58 +01:00
UbitUmarov
a46d6004df reduce ping filter time constant 2014-08-14 21:04:30 +01:00
UbitUmarov
8c657e4837 add a estimator of client ping time, and painfully make it visible in show 
connections console command
 2014-08-14 20:41:36 +01:00
Melanie Thielker
b96f951a6a Merge branch 'ubitworkmaster' 2014-08-13 03:05:54 +02:00
Melanie Thielker
ecae45a21e Revert "make HandlerRegionHandshakeReply processing async and delay it a bit. This" 
This reverts commit 30f00bfb14.

Conflicts:

	OpenSim/Region/ClientStack/Linden/UDP/LLClientView.cs
 2014-08-13 03:05:25 +02:00
UbitUmarov
cf48b814eb remove HandleCompleteMovementIntoRegion delay hack from llUDPserver. If we 
need a delay, we need to do it at end of HandleUseCircuitCode before
feeding pending packets (including that one) into processing queue.
 2014-08-13 01:42:16 +01:00
UbitUmarov
0760fa3106 process AgentUpdates in order with rest of packets. Only give higher priority to chat 2014-08-13 00:17:03 +01:00
UbitUmarov
20b3cab5d1 take agentUpdate checks out of llUDPserver (disabling useless debug) and 
do it only where its supposed to be done..
 2014-08-12 23:57:41 +01:00
UbitUmarov
f811efde85 do the significance test invalidation at MoveAgentToRegion 2014-08-12 22:36:17 +01:00
UbitUmarov
b9224a70c4 no need to go from cos into squared sin when just abs of cos is as good 2014-08-12 22:27:50 +01:00
UbitUmarov
dbbfaf3ac3 force AgentUpdate after CompleteAgentMovement to pass by significance test 2014-08-12 22:17:15 +01:00
UbitUmarov
c0378d5430 actually let delay match throttles as original design 2014-08-10 16:36:20 +01:00
UbitUmarov
ebc18a80bf OutgoingPacketHandler can not be paced by hits on SendPacket() 2014-08-10 14:49:58 +01:00
UbitUmarov
c84a3c3630 also on other case 2014-08-09 16:22:46 +01:00
UbitUmarov
601c50dc1e return even zero costs ( prims on other regions ) 2014-08-09 16:19:07 +01:00