| Commit message (Collapse) | Author | Files | Lines |
|
|
|
|
|
|
|
setting of keepalive is wrong, it should follow the requested one ( or
always as http1.1) only deny if needed (errors). KeepAlive may increase
stress on number of avaiable file descritors.
|
|
inventory folders. The newest version of libomv itself is committed here. Basically, everything that was using the AssetType enum has been combed through; many of those uses were changed to the new FolderType enum.
This means that from now on, [new] root folders have code 8 (FolderType.Root), as the viewers expect, as opposed to 9, which was what we had been doing. Normal folders are as they were, -1. Also now sending folder code 100 for Suitcase folders to viewers, with no filter.
All tests pass, but fingers crossed!
|
|
|
|
is a bad folder. Don't even go to the backend to ask for it, because that will likely kill the sim. Apparently Firestorm requests folder Zero quite often.
|
|
exception. Thanks for the line numbers in the exception trace.
|
|
folders. Also fixing the inventory connector GetMultipleItems, so that if everything is in the cache, it returns successfully rather than unsuccessfully.
|
|
connector. This fixed the problem on my local test grid and generally made things faster. This cache has been needed for a while... there are many parts in the code where the sim gets an item multiple times in a short amount of time (rezzing attachs and objects, for example).
Other minor changes:
- added the scene as a parameter to the constructor od FetchInvDescHandler, so that I could see in which scene the handler was being called
- brought linked items in linked folders back to being prefetched
|
|
folders to go in the reply of the original request. At least my tests indicate that. Pushing this out, so that others who use linked folders a lot more can verify.
|
|
|
|
09:38:40 - [LOGHTTP]: Slow handling of 15572 POST /CAPS/b12c7e98-8261-4953-b7d1-1c414c9893fc FetchInventory2 8acfbca3-13b5-434f-898c-5f4bbe8a76ff from 92.237.199.112:60083 took 62391ms
FetchInventory itself wasn't taking advantage of the new inventory API. This commit fixes that.
|
|
removal. Also added test to check that duplicates are being removed. The test passes. I have no idea how duplicates would not be removed, as reported in the mantis.
|
|
|
|
seems to be a separate issue from the duplicate folders.
|
|
just in case mono has a bug in List<T>.Find, here is the Linq equivalent of distinct-ness.
|
|
|
|
http://wiki.secondlife.com/wiki/Linden_Lab_Official:Inventory_API#Fetch_Inventory_Descendents
Also, consolidated the location of the files that handle inventory capabilities.
|
|
inventory code in place. Cause: # of descendants was wrong.
|
|
inventory links. Added yet another function to IInventoryService to get multiple items at once, so that fetching collections of linked items is done once per folder instead of once per item.
|
|
protocol -- I don't see it here: http://wiki.secondlife.com/wiki/Linden_Lab_Official:Inventory_API#Fetch_inventory_descendents
But who knows!
|
|
request (see http://opensimulator.org/mantis/view.php?id=7054#c28297). This was throwing the new inventory download code off.
Firestorm also sends requests for folder UUID.Zero, which doesn't exist, but doesn't cause any problems either -- it just fails. I'm leaving the warning message for now, just to keep an eye on which folders can't be found, but may remove it in the future.
|
|
related to previous commit.
See http://opensimulator.org/mantis/view.php?id=7054#c28282
|
|
IIventoryService that fetches sets of folders at once. Also added folder id in the InventoryCollection data structure, so that we don't need to go to inventory server again just for that. This reduces the chatter between sims and inventory server by... a lot. On my tests, this reduces initial inventory download down to 30% of what it currently is.
|
|
AssemblyVersion("0.8.2.*")
|
|
it delivers binary and has binary as input.
base64 intermediate coding makes no sense.
Signed-off-by: BlueWall <jamesh@bluewallgroup.com>
|
|
visitor
Signed-off-by: BlueWall <jamesh@bluewallgroup.com>
|
|
that case the response has to report it as bad_folders with uuid instead of returning some dummy
Signed-off-by: BlueWall <jamesh@bluewallgroup.com>
|
|
GetDisplayName cap return data to make the single name appear.
This is still always your avatar name - code to set a different display is not yet implemented.
This works from my testing with current Firestorm and Singuarity releases.
|
|
[ClientStack.LindenCaps] Cap_GetDisplayNames = "localhost" is set by never passing the username."
This reverts commit 2d574c3036964d95dbf914d5f5858a3c39f0c16e.
|
|
first name for display."
Reverting at this point since it may not work well with all viewers.
Requires further investigation.
This reverts commit a29d53986aa3ca9bfd6759789c8a6f434b78d627.
|
|
for display.
This is to match existing behaviour with UDP behaviour where this is masked out.
And to make it consistent with other places where viewers mask this out themselves.
Relates to http://opensimulator.org/mantis/view.php?id=7408
|
|
[ClientStack.LindenCaps] Cap_GetDisplayNames = "localhost" is set by never passing the username.
Unexpectedly, not setting user name achieves this whereas one might have expected it to be done by not setting display_name in the cap OSD.
Emulates LL behaviour when no display name is set.
Relates to http://opensimulator.org/mantis/view.php?id=7408
|
|
|
|
|
|
called WebFetchInventoryDescendents, and we no longer use that cap; the viewers use FetchInvventoryDescendents2.
|
|
names, and this cap returns the regular name. But this moves the server side into the newer, preferred, protocol used by the viewer for fetching the names of agents in the scene given their UUIDs. (the old protocol is via UDP). This works fine in my limited tests, but could use further testing by others.
|
|
To do this required GetMesh to be converted to a BaseStreamHandler
Unlike GetTexture connector, no redirect URL functionality yet (this wasn't present in the first place).
|
|
package rather than some in OpenSim.Tests.Common.Mock
the separate mock package was not useful and was just another using line to always add
|
|
|
|
#Please enter the commit message for your changes. Lines starting
|
|
Stop failure by actually giving the test handler a path instead of null
|
|
This reverts commit f6ea5088f4cc76eddb05b20b5d768f761b0fbd15.
|
|
|
|
|
|
|
|
|
|
use in OpenSim.
|
|
|