[19:47 UTC] *** Logging started. *** [12:47 PM PT] Ubit Umarov: opps [12:47 PM PT] Ubit Umarov: i forgot about the region log :) [12:48 PM PT] Vincent.Sylvester @hg.zetaworlds.com: Well basically what happens is older prims that are in cache in the viewer can be edited just fine from the perspective of the region [12:48 PM PT] Gavin.Hird @grid.xmir.org:8002: now we have to start over again [12:48 PM PT] Ubit Umarov: ok andrew can make a summary of what we said :) [12:48 PM PT] Vincent.Sylvester @hg.zetaworlds.com: But when you force the viewer to redraw the prim it uses the cached version which still shows the previous state of the prim [12:48 PM PT] Gavin.Hird @grid.xmir.org:8002: force the viewer to redraw the rpim? [12:49 PM PT] Vincent.Sylvester @hg.zetaworlds.com: Zoom out real far and back in [12:49 PM PT] Vincent.Sylvester @hg.zetaworlds.com: I posted a video on how to reproduce it to the mantis which shows what happens [12:49 PM PT] Gavin.Hird @grid.xmir.org:8002: how often do you flush changes to persistent storage? [12:50 PM PT] Vincent.Sylvester @hg.zetaworlds.com: On the region it's fine, you edit, that change takes when you nuke the viewer cache it all looks fine, it's definitely local viewer cache messing up [12:50 PM PT] Andrew Hellershanks: ty, Ubit. [12:50 PM PT] Vincent.Sylvester @hg.zetaworlds.com: What's odd is that there seems to be code to tell the viewer to purge the object from cache [12:50 PM PT] Vincent.Sylvester @hg.zetaworlds.com: Let alone the viewer itself is editing the prim so it should know to purge it from cache and retrieve a new copy instead [12:50 PM PT] Gavin.Hird @grid.xmir.org:8002: why is that odd [12:51 PM PT] Gavin.Hird @grid.xmir.org:8002: cache is ment to be purged or invalidated at any time [12:52 PM PT] Vincent.Sylvester @hg.zetaworlds.com: It's probably there to make sure things are purged, but in regards to testing this issue it is strange the viewer doesn't do this itself as well, knowing which prim is being edited so knowing that it shouldn't used the local cache version afterwards yet it still does [12:52 PM PT] Gavin.Hird @grid.xmir.org:8002: if your change has not been written to persistent storage and you do something that force the viewer to invalidate the caceh and retrieve it from persistent storage, you will lose cahnges [12:52 PM PT] Gavin.Hird @grid.xmir.org:8002: so you need to set your region to write your changes back to the db more often [12:53 PM PT] Vincent.Sylvester @hg.zetaworlds.com: The edit on the region works properly that's not the issue, you can verify this using physics [12:53 PM PT] Gavin.Hird @grid.xmir.org:8002: so where does it not work? [12:53 PM PT] Vincent.Sylvester @hg.zetaworlds.com: I scaled a prim down to reveal terrain underneath and when I had the viewer redraw the prim showing the original size the part I scaled back became phantom [12:55 PM PT] Vincent.Sylvester @hg.zetaworlds.com: https://www.youtube.com/watch?v=pYmTDf76jcw [12:56 PM PT] Vincent.Sylvester @hg.zetaworlds.com: Ubit been telling me to leave object cache enabled for performance reasons, though outside of particularly slow regions I don't really see much difference in that, perhaps I am more patient than others when it comes to waiting for a region to load [12:57 PM PT] Vincent.Sylvester @hg.zetaworlds.com: It's just driving me mad when you make changes to something, turn around to have them all undone and you gotta nuke cache and relog each time [12:57 PM PT] Ubit Umarov: i see caching working fine most the time [12:57 PM PT] Gavin.Hird @grid.xmir.org:8002: I don't think ti was undone in your example, because wjhen you walked around there after the zoom, you came to the point where you stepped up to the resized prim [12:58 PM PT] Gavin.Hird @grid.xmir.org:8002: it is a drawing problem in the viewer [12:58 PM PT] Vincent.Sylvester @hg.zetaworlds.com: Yep like I said on the region itself the edit sticks just fine, the viewer just doesn't follow because it grabs the cached version [12:58 PM PT] Ubit Umarov: yeah that clip is irritating ofc [12:58 PM PT] Gavin.Hird @grid.xmir.org:8002: but I don't see the point of those zooms? [12:59 PM PT] Vincent.Sylvester @hg.zetaworlds.com: Forces redrawing [12:59 PM PT] Gavin.Hird @grid.xmir.org:8002: for what purpose? [12:59 PM PT] Vincent.Sylvester @hg.zetaworlds.com: So it triggers cache fetch [12:59 PM PT] Ubit Umarov: and that is not even viewer cache [12:59 PM PT] Gavin.Hird @grid.xmir.org:8002: why? [12:59 PM PT] Ubit Umarov: that is in memory things [12:59 PM PT] Gavin.Hird @grid.xmir.org:8002: exactly Ubit [1:00 PM PT] Vincent.Sylvester @hg.zetaworlds.com: That's the thing, disabling viewer object cache in the viewer directly entirely removes this behavior [1:00 PM PT] Gavin.Hird @grid.xmir.org:8002: why do you want to trigger cache fetch? [1:00 PM PT] Vincent.Sylvester @hg.zetaworlds.com: After doing that I couldn't get a single prim to "misbehave" in that manner [1:00 PM PT] Gavin.Hird @grid.xmir.org:8002: did you only test this in FS? [1:01 PM PT] Gavin.Hird @grid.xmir.org:8002: or does other viewers behave the same [1:01 PM PT] Vincent.Sylvester @hg.zetaworlds.com: I have not tested with other viewers so far [1:01 PM PT] Gavin.Hird @grid.xmir.org:8002: ok, but FS has a very particular cache implementation [1:01 PM PT] Gavin.Hird @grid.xmir.org:8002: so it could be an issue there [1:02 PM PT] Gavin.Hird @grid.xmir.org:8002: also I still don't get the need to trigger a cache feetch [1:02 PM PT] Gavin.Hird @grid.xmir.org:8002: fetch [1:02 PM PT] Vincent.Sylvester @hg.zetaworlds.com: Because when you work on something it randomly does that while you aren't looking and then you turn around to see all the changes you made half an hour ago seemingly undone when they actually aren't [1:03 PM PT] Vincent.Sylvester @hg.zetaworlds.com: Made me question my sanity first few times I saw that [1:03 PM PT] Gavin.Hird @grid.xmir.org:8002: never, ever experienced that [1:03 PM PT] Vincent.Sylvester @hg.zetaworlds.com: The mantis ticket exists simply to track this on OpenSim end and to get verification that OpenSim does send proper information [1:03 PM PT] Gavin.Hird @grid.xmir.org:8002: what is the size of the region you work on? [1:03 PM PT] Vincent.Sylvester @hg.zetaworlds.com: var or normal doesn't matter happens on both [1:04 PM PT] Gavin.Hird @grid.xmir.org:8002: ok [1:04 PM PT] Gavin.Hird @grid.xmir.org:8002: have you tweaked the occlusion settings in your viewer? [1:04 PM PT] Andrew Hellershanks: Allow me to interject a quick reminder. The Open Simulator Community Conference is this weekend. Check out the web site at https://conference.opensimulator.org/ for the schedule of events. [1:04 PM PT] Vincent.Sylvester @hg.zetaworlds.com: Did tell FS devs about this, but haven't heard back, guess I will open a ticket on their end in this regard to get the ball rolling [1:05 PM PT] Vincent.Sylvester @hg.zetaworlds.com: I have ambient occlusion turned off [1:05 PM PT] Gavin.Hird @grid.xmir.org:8002: that was not what I meant [1:06 PM PT] Gavin.Hird @grid.xmir.org:8002: objects off camera can be occluded to speed up rendering [1:06 PM PT] Gavin.Hird @grid.xmir.org:8002: typically things behind you [1:06 PM PT] Gavin.Hird @grid.xmir.org:8002: but also objects hidden by others [1:07 PM PT] Vincent.Sylvester @hg.zetaworlds.com: Outside of disabling object cache I haven't changed viewer debug settings yet [1:07 PM PT] Gavin.Hird @grid.xmir.org:8002: ok [1:08 PM PT] Gavin.Hird @grid.xmir.org:8002: how much GPU memory have you allocated for the viewer? [1:08 PM PT] Vincent.Sylvester @hg.zetaworlds.com: There is a setting for that? [1:09 PM PT] Gavin.Hird @grid.xmir.org:8002: there is for texture memory [1:09 PM PT] Vincent.Sylvester @hg.zetaworlds.com: That's set to 512 mb [1:09 PM PT] Gavin.Hird @grid.xmir.org:8002: ok [1:09 PM PT] Gavin.Hird @grid.xmir.org:8002: how much GPU memory do you ahve? [1:09 PM PT] Vincent.Sylvester @hg.zetaworlds.com: Not sure if that has bearing on it, but worth a try I guess [1:10 PM PT] Vincent.Sylvester @hg.zetaworlds.com: Regular Nvidia 1080 so not all that much and it has to drive 4 screens so probably got 2-4gb left ot play with [1:11 PM PT] Vincent.Sylvester @hg.zetaworlds.com: Like I said merely opened the mantis so we can rule out OpenSim as the cause of this if it does indeed send proper notice to invalidate an object that's been edited [1:11 PM PT] Gavin.Hird @grid.xmir.org:8002: I think I have heard LL say the viewer code is not particularly happy with GPU memory over 4 Gb, but I might be wrong [1:12 PM PT] Ubit Umarov: well opensim did sent right information, you seen prim change size [1:12 PM PT] Ubit Umarov: and the zoom is 100% viewer side [1:12 PM PT] Gavin.Hird @grid.xmir.org:8002: yes it is [1:13 PM PT] Vincent.Sylvester @hg.zetaworlds.com: Yep from going through the support code for object cache I didn't find anything that stuck out to being a potential cause for this if it wasn't working right [1:13 PM PT] Ubit Umarov: so.. vodoo [1:13 PM PT] Vincent.Sylvester @hg.zetaworlds.com: I will make a note to poke FS via a ticket as I haven't gotten a return on my email yet [1:13 PM PT] Ubit Umarov: guess not easy to try repo similar circunstances at sl [1:13 PM PT] Gavin.Hird @grid.xmir.org:8002: I think it might be the viewer does not invalidate the cached item when it is persisted to the region, [1:14 PM PT] Vincent.Sylvester @hg.zetaworlds.com: That would be my guess too [1:14 PM PT] Ubit Umarov: if you relog the prim is as it was edited [1:14 PM PT] Ubit Umarov: no ? [1:14 PM PT] Vincent.Sylvester @hg.zetaworlds.com: Not always, but more likely, if I nuke cache it definitely is as it fetches it from region [1:15 PM PT] Gavin.Hird @grid.xmir.org:8002: but there is caching in the viewer, but also in flotsam, yeah? [1:15 PM PT] Ubit Umarov: so region just did its thing.. [1:15 PM PT] Vincent.Sylvester @hg.zetaworlds.com: I made some experiments leaving cache and just disabling caching system, mixed results there and I don't know the internals of what causes it to invalidate it either [1:15 PM PT] Ubit Umarov: nothing to so with flotsam [1:15 PM PT] Ubit Umarov: prims are not even on it [1:15 PM PT] Gavin.Hird @grid.xmir.org:8002: ok [1:16 PM PT] Vincent.Sylvester @hg.zetaworlds.com: This issue has been around for years now, I have seen that first time all the way back in 2017 even [1:16 PM PT] Ubit Umarov: ( unless if rez from inventory possible ) [1:17 PM PT] Vincent.Sylvester @hg.zetaworlds.com: It must be something small just a missing line somewhere in the viewer code, but chances of me finding that within the decade are slim [1:17 PM PT] Gavin.Hird @grid.xmir.org:8002: but in any case it should be invalidated in the viewer cache when persisted to the region [1:17 PM PT] Ubit Umarov: i remember there was a viewer cache option that did show prims, if we did move away and back.. boing all gone till relog :) [1:18 PM PT] Ubit Umarov: i show a fs dev.. answer was " why did you set that option.. it is not used.." [1:18 PM PT] Ubit Umarov: done :) [1:18 PM PT] Andrew Hellershanks: As you are talking about viewer behaviour and caching I'll add in a comment/question in case it might be related. [1:19 PM PT] Andrew Hellershanks: I was noticing the other day that after I logged in to a region some of the textures were blurry (because they had not fully loaded yet?). I forced a texture refresh and they fully loaded and looked good. A few moments later they went blurry again then eventually fully loaded once again and looked ok. [1:20 PM PT] Andrew Hellershanks: It seems a bit odd that a fully texture would "unload" and need to be loaded a second time. [1:20 PM PT] Gavin.Hird @grid.xmir.org:8002: texture trashing [1:20 PM PT] Gavin.Hird @grid.xmir.org:8002: too little texture meory most likely [1:21 PM PT] Vincent.Sylvester @hg.zetaworlds.com: Dayturn used to do that for me quite a bit, few releases ago it stopped though [1:21 PM PT] Andrew Hellershanks: I hadn't moved. It was like a given texture wasn't loaded fully, then loaded and showed properly then reverted to a partially loaded state, then settled down to a fully loaded state. [1:22 PM PT] Andrew Hellershanks: The only thing I was doing during that time was using Tex Refresh in the viewer to help it load a partially loaded texture. [1:22 PM PT] Vincent.Sylvester @hg.zetaworlds.com: Viewer loads close things first then everything farther away so it might eventually load a prim the next region over causing it to fill texture memory [1:22 PM PT] Gavin.Hird @grid.xmir.org:8002: There has a been a lot of fixes to texture memory allocation, so that basically got rid of it [1:23 PM PT] Andrew Hellershanks: Vincent, ok. Sounds like it viewer side. I'm running the latest version of FS. [1:23 PM PT] Gavin.Hird @grid.xmir.org:8002: plus if you have oversize textures, meaning larger than 1024x1024 it goes nuts about it [1:23 PM PT] Ubit Umarov: just see it happen [1:23 PM PT] Ubit Umarov: if a prim is a old to be cachable.. [1:23 PM PT] Andrew Hellershanks: Wasn't my region or my objects so I have no idea of the size of the texture(s). [1:24 PM PT] Ubit Umarov: ( simple box) [1:24 PM PT] Ubit Umarov: we log in.. resize it.. zoom out completly that in.. the size change is gone [1:24 PM PT] Gavin.Hird @grid.xmir.org:8002: older objects might have 2048x2048 textures applied to them [1:24 PM PT] Ubit Umarov: fs shoes the prim as it was at login even on edit tab [1:25 PM PT] Ubit Umarov: we relog, and go get the changed prim [1:25 PM PT] Ubit Umarov: so.. fs is doing some mess [1:25 PM PT] Ubit Umarov: i can't test at sl [1:25 PM PT] Gavin.Hird @grid.xmir.org:8002: it is still possible to do that, but you have to sneak the texture that size in via direct upload to db [1:25 PM PT] Ubit Umarov: bc can't rez a box for 30 mins or to [1:25 PM PT] Ubit Umarov: or so.. [1:26 PM PT] Ubit Umarov: test at sl rez a box and leave there for some time ( 30min) ? [1:26 PM PT] Ubit Umarov: so it is flagged as cachable [1:27 PM PT] Ubit Umarov: then login and repeat change size zoom out, zoom in.. [1:29 PM PT] Andrew Hellershanks: We are at about the hour and a half mark. Does anyone have any other questions before this meetings is wrapped up? [1:30 PM PT] Andrew Hellershanks: I don't see any one typing so I will take that as that is all for today. [1:30 PM PT] Selby.Evans @grid.kitely.com:8002: https://www.reddit.com/r/metaverse/ [1:30 PM PT] Andrew Hellershanks: I hope to see some of you at the OSCC event this weekend. [1:30 PM PT] Jamie.Jordan @grid.kitely.com:8002: great meeting [1:31 PM PT] Andrew Hellershanks: Enjoy the OSCC events. We can talk about it this coming Tuesday. [1:31 PM PT] Andrew Hellershanks: Thank you all for coming. See you again next wee.k [1:31 PM PT] Gavin.Hird @grid.xmir.org:8002: Thanks Andrew [1:31 PM PT] Andrew Hellershanks: yw, Gavin. [20:32 UTC] *** Logging stopped. ***