aboutsummaryrefslogtreecommitdiffstatshomepage
Commit message (Collapse)AuthorAgeFilesLines
* make oar/iar assets writer be done by caller thread like the rest of theUbitUmarov2017-06-213-176/+50
| | | | | | oar/iar save. This may look more painfull but should reduce errors and threads fights. Fill lucky i just don't stop the entire simulation during this
* changes to AutoBackModule. Store folder and number of days expire is now ↵UbitUmarov2017-06-212-133/+89
| | | | also only defined in OpenSim.ini and so same for all regions.
* changes to AutoBackModule. Add option ALL to dooarbackup to save all ↵UbitUmarov2017-06-211-2/+14
| | | | regions in instance
* changes to AutoBackModule. Time interval is now unique and only definable ↵UbitUmarov2017-06-212-556/+221
| | | | on OpenSim.ini. All enabled regions will be saved in sequence whne it expires. Interval is time since last region save end, so will not follow wall clock. This reduces the chance of overlapped saves. Console command renamed to dooarbackup to be more distint from db backup. Lost the region load checks, to put back later
* revert EnvironmentTick back to orignal clock, since change may cause issues ↵UbitUmarov2017-06-193-84/+49
| | | | on some code paths. Clean a bit get mesh and get texture throttle
* a defual can me comented outUbitUmarov2017-06-191-1/+1
|
* framework main thread pool is always active and in use ( even id hard to ↵UbitUmarov2017-06-163-24/+47
| | | | catch) so show in on show stats. Disable ServerStatsCollector by default, since most don't use it, Adicionally it uses shared framework performance counters system that may be affected if a region crashs
* change the clock source for EnvironmentTickCount so it does get a bit more ↵UbitUmarov2017-06-151-8/+10
| | | | resolution if avaiable (1ms) specially on windows. This until all calls to this are removed. Coment out some stats in workpool/threads creation path
* update inventory item permissions on deattachUbitUmarov2017-06-151-0/+25
|
* fix llsd DeserialiseOSDMap cast from boolean to int, this seems to fix ↵UbitUmarov2017-06-141-0/+5
| | | | alchemy beta 5.0 scripts upload, but this viewer seems to be of no use for opensim, sice it does crash.. insists lbsa is at second life etc
* some cleanup.. remove some stats that only some do look at once on a lifetimeUbitUmarov2017-06-145-181/+13
|
* remove options about script compilers the got back somehow on defaults alsoUbitUmarov2017-06-141-8/+0
|
* remove options about script compilers the got back somehowUbitUmarov2017-06-141-11/+0
|
* main generic use JobEngine also does not need a permanent thread.. actually ↵UbitUmarov2017-06-141-1/+1
| | | | doesn't even seem to be in use
* guess HGEntityTransferModule jobengine does not need a permanet threadUbitUmarov2017-06-131-1/+1
|
* coment out tests that now depend on active thread poolUbitUmarov2017-06-131-1/+2
|
* change llclient async udp packets processing. Removed the shared Jobengine ↵UbitUmarov2017-06-132-11/+15
| | | | that made only one thread do some packets for all users, and removed the FireAndForget, so each user could issue a unlimited number of threads. Added a new JobEngine per client, so each gets at most one thread,and that thread can be released if iddle (for 5 seconds )
* make JobEngine be a workitem of mail pool (smartThread), with the option to ↵UbitUmarov2017-06-131-48/+41
| | | | release thread after a idle time, so is free to do other service elsewhere
* add option for Fireandforget not timeout (our access to main smartThreadPool)UbitUmarov2017-06-132-9/+9
|
* hope gc does remove a little dic and contentsUbitUmarov2017-06-131-5/+3
|
* dont let time go backUbitUmarov2017-06-131-7/+31
|
* some cleanupUbitUmarov2017-06-134-36/+16
|
* udp is not tcp. If mono versions have bronke udp sento, better update, also ↵UbitUmarov2017-06-133-5/+27
| | | | having 300 threads because object select is not funny
* remove excessive notion of paralelismUbitUmarov2017-06-133-31/+5
|
* update Robust32.exe build filesUbitUmarov2017-06-129-27/+200
|
* add Robust32.exe so recovering the option to run it also in 32bit on windowsUbitUmarov2017-06-125-18/+99
|
* fix some issues on ubOde physics shape type changesUbitUmarov2017-06-122-15/+18
|
* improve english on few messages (thx)UbitUmarov2017-06-122-3/+3
|
* make some web request errors visible, so cause of later asset not found is clearUbitUmarov2017-06-121-7/+13
|
* try to mk jenkins happyUbitUmarov2017-06-121-1/+1
|
* stop doing UpdateMovementAnimations on avatar collisions, that may cause ↵UbitUmarov2017-06-121-2/+2
| | | | incorrect changes
* add/replace sources of OpenSim32.exeUbitUmarov2017-06-117-26/+213
|
* fix a error messageUbitUmarov2017-06-111-2/+5
|
* forget about stupid mono 32bitUbitUmarov2017-06-111-5/+0
|
* actually.. its opensim, use same log file name in 32b modeUbitUmarov2017-06-111-2/+2
|
* remove a ref to old 32bitlauncherUbitUmarov2017-06-111-2/+2
|
* Set the execute bit on the bin/opensim*.sh filesKevin Cozens2017-06-112-0/+0
|
* add opensim.sh and opensim32.sh for linux, replacing old opensim-ode.shUbitUmarov2017-06-113-4/+10
|
* put back opensim 32bit launcher, renamed as OpenSim32.exe. This allows ↵UbitUmarov2017-06-114-75/+75
| | | | opensim to run in 32bit mode on 64bit machines. Possible most opensim regions should be running like this. 2GB of Virtual memory is still a lot on memory
* add a adicional in transit flag to signal HG tps, and use it to ignore ↵UbitUmarov2017-06-096-6/+33
| | | | usernames requests sent to start region during tp; don't send unknows display names ( getdisplaynames cap )
* only silent remove threads from watch list if they stopped ( ie still ↵UbitUmarov2017-06-091-4/+2
| | | | consider aborted etc )
* make SendCoarseLocations async, since it is http. Make some actions use ↵UbitUmarov2017-06-092-13/+20
| | | | thread from pool and not a new one. Threading does need a deep cleanup one of this days. This stops mantis 8183 warnings, but as side effect only
* Minor changes to commented text in bin/OpenSim.ini.exampleKevin Cozens2017-06-091-9/+8
|
* Fixed typo in log messageKevin Cozens2017-06-091-1/+1
|
* still losing time with of halcyon incompatibile oars... cathch exceptions of ↵UbitUmarov2017-06-091-8/+6
| | | | the function that actually throws; Reduce log spam; don't let a broken MOAP stop all object deserialization. Fixing MOAP does not mean halcyon oars issues are fixed, just one. And really we should not even try to go against halcyon decison to be incompatible.(our MOAP encoding did not change since 2010)
* Another attempt at parsing MOAP <Media> elements in OAR files.Robert Adams2017-06-081-1/+22
| | | | | | | | | | Seems there are multiple interpretations of the format of the content of the <Media> element in OAR files. OpenSimulator (for reasons lost in the mist of time) escapes the XML in the element and then reparses it was a separate XmlReader. Other simulators fill the <Media> element with regular XML. This patch parses the <Media> escaped XML content as it always has and, if the parsing fails, falls back to trying to parse the pure XML.
* bug fix plus some cleanupUbitUmarov2017-06-091-56/+65
|
* keep the reverted code, that does work. Our code likes to have pbs.Media == ↵UbitUmarov2017-06-081-15/+28
| | | | null when there is no MOAD defined, so handle possible odd oars that may have llsd <array /> on that case
* revert fa5bf4fd0bb6a855eacdb7b5eec9cd71ad9bf606 for testUbitUmarov2017-06-081-2/+1
|
* so no real need for Maintenance thread.. so removedUbitUmarov2017-06-071-77/+1
|
generated by cgit v1.1 at 2024-11-06 08:37:16 +0000