Deploys for the week of 2013-04-01

[Malate Banufong]

Maestro - Thanks to everyone at LL for fixing the sim crossing bug. It is brilliant that we can fly again. I am loving it.

[Maestro Linden]

---

Norton Burns wrote:

This is hugely unscientific & merely based on hearsay & poke-it-and-see...

... but flicking alpha masks off & on draws everything in straight away when faced with this situation, in my experience.

---

I see similar behavior when I have a live case of this bug.  I found that toggling 'Atmospheric Shaders' off and then back on in graphics preferences made the missing prims appear.

[WolfBaginski Bearsfoot]

And yet on the tests I have done so far, the missing prims I have seen have been parts of linksets, present 9in the physics model but invisible. It was all fitting in a pattern, and then your edit popped up.

I use Firestorm, and it hasn'r had a release since December, so I was willing to write it off as a Viewer problem. I know of a couple of glitches glitches that are gone in more recent viewers.

Thanks for keeping us informed.

Vehicles are much better. There is a pirate battle underway. There has been a rush to get back into the skies over Second Life. The last time I saw this much activity at an airfield they were scrambling nuclear bombers.

 

[Eric Castanea]

---

Maestro Linden wrote:

---

Norton Burns wrote:

This is hugely unscientific & merely based on hearsay & poke-it-and-see...

... but flicking alpha masks off & on draws everything in straight away when faced with this situation, in my experience.

---

I see similar behavior when I have a live case of this bug.  I found that toggling 'Atmospheric Shaders' off and then back on in graphics preferences made the missing prims appear.

---

I just tried that, can confirm.

[Amber Murfin]

I can confirm, the missing prims I posted above are a mixture of all 3 physics types (None, Convex Hull and Prim)

 

Edit to add, this is some of the Best sim crossing I have ever seen in SL for sailing. Thank you

[tristangarrard]

Something I noticed tonight, and this seems to happen on multiple sims, is that the textures will complete loading and the viewer stats will show a bandwidth of circa 10kb/s as normal, but my firewall (Little Snitch) still shows data constantly downloading from the sim IP address at a rate of about 150kb/s.  This does cause some inevitable lagging as well as hammering my bandwidth, I think I saw about 215MB come down in total from the sim before I relogged.  Relogging seems to clear it for a bit but then it restarts.

Anyone else seeing this?

[Theresa Tennyson]

Sim crossings do seem to be much better but on my motorcycle (a notoriously bad sim crosser) I'm hitting a few rough patches and upon checking they seem to correspond to crossing in or out of LeTigre/BlueSteel regions. Could you confirm that those channels got the interest list changes that went out to the rest of the grid?

[Marianne McCann]

---

Chriss Rosca wrote:

This is not something new, it is happening for about 8 month now i think. It is not reproducable, but some places suffer it more than others. I am not sure it is a server issue. When you right-click the wall ot floor it is rezzed so fast, that all the information must have been in the viewer. For some reason however the viewer does not rez the prims until they are touched.

---

Ya, I've noticed this for about the same amount of time, vbut it really is becoming more and more commonplace, even on a fairly decent rig and connection

[Whirly Fizzle]

---

Maestro Linden wrote:

---

Norton Burns wrote:

This is hugely unscientific & merely based on hearsay & poke-it-and-see...

... but flicking alpha masks off & on draws everything in straight away when faced with this situation, in my experience.

---

I see similar behavior when I have a live case of this bug.  I found that toggling 'Atmospheric Shaders' off and then back on in graphics preferences made the missing prims appear.

---

Confirmed flipping alpha masks and atmospheric shaders settings brings all the missing prims back.

So does toggling "Bump mapping and shiny" setting.

Zooming camera far away and back also works.

 

[Monty Linden]

---

tristangarrard wrote:

Something I noticed tonight, and this seems to happen on multiple sims, is that the textures will complete loading and the viewer stats will show a bandwidth of circa 10kb/s as normal, but my firewall (Little Snitch) still shows data constantly downloading from the sim IP address at a rate of about 150kb/s.  This does cause some inevitable lagging as well as hammering my bandwidth, I think I saw about 215MB come down in total from the sim before I relogged.  Relogging seems to clear it for a bit but then it restarts.

Anyone else seeing this?

---

The Texture Console speaks truth for texture fetches, either http or udp.  If that is quiet while this transport is going on, it's something else.  Little Snitch can tell you more and here are some rules that will determine the traffic:

• Port 12046 but textures are quiet => mesh fetches
• Port 12043 (corrected, was 12042) => other HTTP services ("Capabilities")
• UDP port 12035, 13000-130XX => simulator communications

 

[Danny Nolan]

I have a similar problem that you could look into. When you're experiencing this ghost bandwidth consumption, open up your log in SL and see if you're seeing this problem: http://jira.phoenixviewer.com/browse/FIRE-8842

In summary, for me being on a slower connection, my client will attempt to load 32 (default "meshmaxconcurrentthreads") mesh items at the same time, and on a slower 1.5mbps connection, getting all 32 transfers to complete in under 30 seconds is impossible. So what happens for me, is I get these monsterous amounts of threads downloading mesh, timing out, then starting over - which will literally continue to happen forever until I relog.

On my jira post, I've posted a very - very simplistic fix that should completely alleviate the issue.

Edit: Changed link to firestorm's JIRA.

[Ciaran Laval]

Only a few people will be able to see your Jira Danny, most of us will get a permission violation due to how the Jira is setup these days.

Was the fix a viewer fix or a patch or something else?

[Danny Nolan]

Oh that's right. I forget about that. I mirrored the JIRA on firestorm's JIRA. Should be able to see this one: http://jira.phoenixviewer.com/browse/FIRE-8842

[Darien Caldwell]

It's also possible some of the 'missing prim' reports are due to the packet loss that was going on. It was pretty bad.

 

However, I can report at least for my neck of the woods, the packet loss has finally subsided. If LL did anything to correct that, thanks.  0% packet loss, no "invalid objects' = Happy camper.

[Ayesha Askham]

@Dari

Never saw the packet-loss issue on Woods of Heaven sim (also Letiggywinkle), but the "missing prim" bug is very present!

I find that the toggle on alpha masking in Prefs>Graphics works, but of course you have to do it every TP, even in-sim.  Still it IS a workaround!

 

@Maestro

Any news from the viewer team as to what they'll do to fix this?

[Domitan Redenblack]

Yes, I am seeing Interest List failures after every tp now. Much worse than before, and it was not great before.

 

 

[WolfBaginski Bearsfoot]

Vehicle use is generally good. Some of the old problems are still there, but we're back at the quality level that followed the intr#oduction of the threaded code, Still glitches apparent at the moment of region crossing, but few total failures,

On the missing prim problem, while I usually hear people warning against cache clearing, I chose to clear cache, going by the possibility that there might be bad data which explains why particular prims always seem to be invisible. And, so far, that does seem to have made a difference. I'm not sure this is a good idea, but if you're seeing the same prim go missing every time, it may be worth doing.

Clearing Cache is not a panacea, and your bandwidth consumption and general lag will go up for quite a while.

[Perrie Juran]

@ Maestro

I too have been having the 'missing prim' problem for a while.

One thing I have been noting is that often times as I move about SL it is always the same prim at a location that doesn't show.  A wall at this place, a floor at another, a door on my house.  Generally simply clicking causes it to appear.  I am thinking there must be a common denominator but I have not been able to determine one. I simply don't have the technical expertise to figure this stuff out.

[Ayesha Askham]

@Wolf

I am very wary of recommending a cache clearance as a remedy for anything other than a texture corruption.  It may work, but I would doubt very much whether it would be effective for everyone.  Apart from any other considerations it is extremely time consuming, and a TP will render (if you'll pardon the pun) the operation ineffective.

That the render options are effective until you TP is, I think a sign that communication from sim to viewer is now in some way broken, and that really is the province of Linden Lab to sort out a "fix".

To Simon Linden, and the others on the "Interest List" programme, I say this (though it is highly unlikely that you'll see it):

This work was conceived in good faith and with laudable aims.  However, it is now time to recognise that it is ineffective in its objective of improving the "shared experience" and appears to break more than it fixes.  This is the time to call a halt to it.

[Qie Niangao]

As someone who has had the "missing prim" problem for months, I want to add that for me at least, it's no worse now than it was before. In fact, exactly the same prim most reliably goes missing as it has for all those months.

I suspect it is related to another case of stuff going invisible, reported a couple weeks ago in a scripting forum thread. That's an interesting case because it seems to be Pathfinding-related (I've only seen it on llCreateCharacter() prims*), and although it takes a few minutes patience before it actually kicks in, the phenomenon is conveniently repeatable, at least for me.

That's not to say that any of this is Pathfinding-related, per se. That "most reliably missing" prim I mentioned is certainly not a Pathfinding character and neither are the other items mentioned here. I did, however, first notice the missing prim problem around the time Pathfinding was introduced, so I'm thinking that the viewer was relying on something that the sim sometimes omits, since about that time. If it's true that, in general, it happens more now, it may be that the Interest List changes simply remove some of the update redundancy that used to hide the underlying viewer problem.

__________

*Originally I saw it with a regular box prim, but at the moment I'm only reproducing it on a Mesh; I'm not sure if that's a matter of patience or if something is different.

[Cincia Singh]

I too experience prims that are invisible until I click on them, and clearing cache does reduce the incidence temporarily. In fact clearing cache from time to time is a good thing, just like restarting a computer periodically is a good thing; the reasons for this are too numerous to state here. I seriously doubt clearing cache will resolve the invisible prim issue, but clearing cache from time to time is still a good thing. And there's no reason to "call a halt" to anything. Move on and improve, don't regress and stagnate. The current state of SL is better than it's ever been (at least since 2007) and it'll only get better as features and improvements roll out!

[Norton Burns]

"but clearing cache from time to time is still a good thing."

 

Why, oh why, oh why, do people still maintain this?

Cache is like soap... self-cleaning.

Leave it alone unless you have good reason to clear it. It's not like an engine, it doesn't need an oil change every 10,000 rezzes.

Read http://wiki.phoenixviewer.com/cache for a good idea of when to & when not to blame cache for issues.

[Ayesha Askham]

Cincia

Leaving aside the cache clearance debate - Norton says enough - perhaps "calling a halt" would indeed be too extreme.  Nonetheless at the least a moratorium needs to be called until the deleterious effects upon the viewer code have been assessed and countered.

The "Shared Experience" has always been the Lab's prime objective and anything that worsens it is anathema to them.  There can, I would have thought, be little more destructive to the "shared experience"...to any level of experience of the virtual world of SecondLife, than having segments of it fail to render!

To call it a "Gamebreaker" would be understandable but possibly extreme, but failure (i.e. non-appearance) of key elements of a game would surely break it?

Stopping doing something that is obviously counter-productive is not stagnation, Cincia, it is common-sense!  I do not advocate SL regressing or stagnating, I merely would like it not to break.

 

[WolfBaginski Bearsfoot]

The self-cleaning nature of the cache depends on how big it is, and your pattern of using SL. 

I was seeing consistent prim errors in regularly visited locations. Those particular errors stopped when I cleared cache. It hasn't cured the underlying problem, and I wouldn't want to do it again any time soon, but I reckon I made a good choice, this time. I saw a positive effect.

Note to Maestro: when the fix is fully deployed, I shall probably clear cache again. But that does depend on a clear statement of what changes are in the code. I don't think in terms of JIRA numbers.

[WolfBaginski Bearsfoot]

In my experience, camera zooming is not always effective. I note that the Interest List code has had a revision that suggests time could be a factor.

Since it's been stated that gadgets such as the graphics-cleaning rockets should be modified to hold the AV at altitude for at least five seconds, waiting that long before zooming back seems like a good idea.