Rolling restart for 1.24.7 2008 Sep 30-Oct 2

Leonardo : I'm not aware of a crash-when-teleporting-after-using-search bug. (That doesn't mean it doesn't exist.) What I'd suggest is that you search the JIRA to see if there is a bug submitted on it already (search for 'crash search teleport' or some such). If you don't see it, submit a new JIRA with as much information as you can. Be aware that for us to do anything about a JIRA issue, we need to be able to reproduce it. That means that you need to give as detailed and specific of information as possible.

See http://wiki.secondlife.com/wiki/Public_Nightly .

I think the origin of the name comes from Sherlock Holmes..

lol have you seen the hundred's that are well past 120 days old - collecting dust - totally untouched?

Okay, same boring question, has the latest rolling restart finished? That's the one to 1.24.9. And I guess the last sentence was just copied from the previous Restart - "Please watch the Status Blog for detailed information, which will be posted after the Rolling Restart has completed."

That's depressing.

It totally waves a white flag over anything that's timing-dependent or sensitive to the environment.

I'm all for reporting all the available information, and if something *is* solidly reproducible then that's great, and makes the the job of the dev assigned to the issue much easier.

But some of the most vexing bugs today defy easy reproducability...viewer memory leaks, for example. These will require either static analysis of the source code to inspect it for memory management errors, better diagnostic tools, or both.

"We can't fix it if the user can't tell us exactly how to reproduce it" isn't good enough for that kind of problem on a system of this complexity. Iif you need better descriptions of what the viewer was doing at the time of the crash and just before, then the viewer code itself is going to have to capture it, and transmit it as part of the crashdump (rather than asking an end-user to remember everything that was going on from a technical point of view at the time when they were rudely ejected from their immersive experience.

This may take longer than what you're doing now. Most users wouldn't care about that *if you let them back in-world while the dump was in progress*. And when a correctly authenticated user tries to reconnect to the server they were on when the viewer failed on them, they deserve better treatment than "come back later after your session times out".

In fact, a crashed viewer *could* schedule a comprehensive dump in the background, and automatically log the user back in again. This alone would take a lot of the pain out of instability.

This is just one more example of the total inepetitude at communications and customer relations that has become a hallmark of LL®. Even a simple, "this is why we are doing this and more details will be available soon", would have been preferable to this total lack of respect that customers of LL® are being shown.

Everyone working at LL® needs a good lesson in business communications and if the so called "Communication Linden" is still working there she should be terminated immediately for incompetance. I'm sorry if my words seem harse but poor communications is a business killer and I would prefer that LL® and SL® not die.

Please, I can't give any more information about the rolling restarts just now. Things will be explained when we are able to explain them. Watch the status blog for more information.

Meh, when I'm working on a product in-world and send it out to my beta-testers, I tell them that reporting bugs to me is only useful if they report it in such a way that it's reproducible. Just the other day, two of my beta testers reported a fairly minor (but important to have fixed) bug involving muzzleflash going off when it shouldn't. Both gave incomplete summaries of what they did to get it the bug to occur, and since I couldn't make it happen, I couldn't fix it....until one of them finally said "oh, I'm using the XXX bullet, does that make a difference?" /me facepalms. Long story short, yes it did. Once I finally had a full reproduction, I was able to fix the bug in 30 seconds.

My point as it applies to SecondLife as a whole is fairly similar. Programmers need to be able to reproduce a problem in order to figure out the most basic things such as where in the code it's occuring. The larger the codebase, the more essential this becomes. In my example, I suppose if worst came to worse, there were only a couple of scripts with a couple hundred lines of code each to look through if I'd had to go line by line. With SecondLife, that's well nigh impossible. Without a solid repro, it's not even obvious whether the problem is client-side or server-side. LL can't even assign it to the proper people to look into it without a reproduction, since the repro is what tells you where to look. SecondLife is a large and complex program, and I highly doubt that there is a single person at LL who is familiar with every part of it...even if there were, the majority of the staff are assigned to and familiar with a specific area of the program.

And of course, don't forget that asking LL to try to find a repro themselves would mean taking productive time away from working on solving problems for which a solid repro exists and thus a potential solution would be much easier to find. I agree that for a few clientside errors, solid repros are hard for the user to create, but by and large it shouldn't be too difficult to come up with a reproduction for most serious problems. All it amounts to is "every time I do X with setting Y, Z happens." Then people can test it, and if they can reproduce Z by doing X with setting Y, then they can delve into the code and figure out exactly what doing X causes, and how setting Y affects it. Note that often, if you simply say "When I do X, Z happens" might not be reproducible if the person trying to reproduce it doesn't know that they need to have setting Y. Without that critical piece of knowledge, the problem might be impossible to fix.

LL engineers have a finite amount of time and resources with which to fix problems. Asking them to try to reproduce bugs on their own because some users are too lazy to do it themselves would simply mean that fewer bugs would be fixed.

Has it not occurred to anyone that there may be valid legal and business reasons/considerations why we're not being told what the underlying issue is? I can think of several good reasons to not tell "me," and I'm sure someone more knowledgeable about MMO's and the technology involved in SL could think of many more.

Sure it has!

But there are some people would rather see _any_ action or inaction or not-sure-if-that-was-action-or-not as proof that LL is being stupid or that they hate us residents or etc, etc, blah, blah, blah, whine, whine, whine.. Upside is that they're pretty easy to spot and then ignore.

Sounds like a "when we get out a new required viewer that plugs a dangerous exploit and now nobody can access it we will tell you what it was"

No it means that there is something CRITICAL that was found and they don't want to give ANYONE, ANY HINT as to was it it, least we have a herd of griefers trying to exploit it and make it worse.

Which would you rather have, a 10 minute restart, or your sims going down every few minutes as they were griefed out of existence??

Let's have a little common sense here people.

I have found, that common sense on this forum is few and far between.

As if the griefers wouldn't figure it out from the fake Navy Seal "I could tell you but then I would have to kill you" responses we have been getting? How about a simple "A problem has been found in the latest update requiring an immediate fix. More details will be available once we have verified the new update."? That at least sounds like they are fixing a problem and not trying to hide something from us. And they have said in the past that there have been security problems without causing any griefing problems. They do not have to give details of the problem before it's fixed but the cloak and dagger BS they are feeding us promotes speculation and anger. The key is good communications, something that LL® has been doing a very poor job of. A simple "A problem has been found and we are fixing it." would be a lot better than just starting a rolling restart, in this case two restarts with no communications at all. Please note that I am complaining about the typical lack of effective communications, not the restarts themselves. I understand that sometimes fixes like this are necessary but that does not relieve them of a duty to keep their paying customers, and some of us do pay and pay alot, informed.

OK, if that's your logic. WHAT IS THE PROBLEM? Explain it please? I hope you're smarter than the average griefer.. you said "as if the griefers wouldn't figure it out"... ok, you've see all the information what would have? WHAT IS THE PROBLEM? Explain in detail. Answers on a postcard...


(aside: Chandra, you're right)

That's all well and good for one of your little LSL-scripted guns.

A massive client/server app like SL is another beast altogether. The difference in scale between the two is *huge*; different techniques are required.

Let's see, LL® gets all secretive about the reason for the restrats. What's the first assumption from everyone? That there is a security problem. So saying they are doing a restart to fix a security problem tells the griefers nothing that they have not already figured out.

As to the problem? Read the last sentance of my previous post, if you can't figure it out from there you never will.

Poor communications kills companies.

i think what really kills these forums is the battles with tons of posts from no nothing linden appollogists, who defend the lindens action with no knowledge of what they are actually doing (unless you are a linden, ilana), and the trolls that just have to have a problem with everything.
i happen to agree with ravanne. this latest series of restarts are an example of terrible communication skills by the lindens. theyve shown us in that past how it relly should be done- explain the problem, what they did to fix it, why that didnt work and what they didnt subsequently. they havent done the minimum on this one, which is tell us that they are doing yet another restart until they are hours along. it really is bad business.

sl is my job. i have a shop here, and im a developer, so sl buys my food and pays my rent. i want nothing but the best for sl and the lindens, cuz frankly i really like eating and not being homeless. i think intelligent informed constructive critisism could be invaluable to the lindens. IF THEY COULD FIND ANY OF IT BURIED IN ALL THE FLAMES going on in the forums. if i was a linden, i wouldnt even want to read the forums.

"can't tell you" can mean "I don't know" it can also mean "I'm not allowed"

Maybe Prospero LIKES his paycheck.. and wants to keep it.

Hey guys -- the rolls are done, I'm going to close off this forum thread. It's devolved into flamewars and attacks, anyway.

There is another forum thread about the security vulnerability. Please try to keep it civil and constructive over there.