Daikon Forge

Archive Page 4 of 65

Taketori Katana V1.0.8 Beta IS NOT USING MONO

Sep
03
2008 in JIRA, Rants, Bug Reports and C:SI.

I just wanted to point out that the rumors that the Taketori V1.0.8 RC2 Beta is using Mono-compiled scripts are false.

I did allow a few people to test a Mono version of the Taketori, and that fact seems to have become confused with the fact that the recent RC2 beta is intended to solve the same problem.

The difference is that the RC2 beta attempts to solve that problem (SVC-2958) WITHOUT using Mono.

Please inform anyone you know still propagating this rumor that it is not correct. I would very much appreciate it

[UPDATE]

Since Linden Lab is now deploying the fix for the state_exit() bug mentioned above, there does not seem to be any more need for this beta, whose purpose was primarily to ensure that people had a weapon that they could use until a fix was deployed :

Update 2008-09-03 05:37am : The first-half rolling restart is in progress. Odd-numbered hosts that do not yet have 1.24.4 are being restarted this morning.

Update 2008-09-02 09:37pm : The pilot roll is complete.

Update 2008-09-02 08:47pm : the pilot roll to 2859 regions has begun.

We will be performing a rolling restart to deploy server version 1.24.4 to Second Life. This server deploy does not contain any new features, but has fixes for some bugs found in the deploy of server version 1.24 that was completed last week.

Release notes for version 1.24 (and 1.24.4) can be found here on the Second Life Wiki.

The rolling restart will follow our usual three-stage deploy:

Tuesday, 09/02, 08:00PM : A pilot group of ~3000 regions will receive 1.24.4 in a rolling restart.

Wednesday, 09/02, 05:00-09:00AM : Half of the grid will receive 1.24.4 in a rolling restart.

Wednesday, 09/02, 07:00-11:00PM : The remainder of Second Life will receive 1.24.4 in a rolling restart.

As is usual with rolling restarts, regions will be down typically for 5-10 minutes. If your regions stays down for more than 20 minutes, please contact support. There is no way to delay the restart of a region once the rolling restart has begun.

Discussion of this rolling restart will happen in the Second Life forums here.

Because the Beta is no longer needed, it is no longer available. I’m relieved to hear that a resolution is underway, and quite frankly the process of creating and distributing a ’stop gap’ Beta was every bit as much of a headache as I had anticipated

Popularity: 37% [?]

Share This

Mono servers causing severe delays in old LSL scripts

Aug
30
2008 in JIRA, Mono, Bug Reports and C:SI.

35 Comments

I’ve complained about the symptoms, but now we may know the cause. Thanks to some sleuthing by Makkiavelli Toland and my own experiences in trying to create a halfway decent workaround, it seems extremely likely that the cause is in the state_exit event being delayed.

Tip of the samurai helmet to Escort DeFarge, who notified me that there is a JIRA issue already on file for this: SVC-2958. Please go vote, and make your voices heard! We need Linden Lab to address this issue.

If history is any indication, it will take Linden Lab some time to acknowledge, research, and distribute a fix for this, so I will continue working on a workaround, and will distribute it as soon as testing is complete. I’d rather have them do it right than distribute another batch of bugs

Also, remember that Archanox is working on a long-term plan to move the C:SI scripts to Mono, and will almost certainly do a better job of getting things working than my hackish workarounds, and I’ll share anything I discover with him.

I am relieved. Time for a nap

[UPDATE]

Periapse Linden commented on SVC-2958:

Babbage has checked in a fix for this. Vektor is testing the new version now. We intend to complete a rolling restart this week with this fix and fixes for several other serious issues.

Popularity: 54% [?]

Share This

Mono is teh suxxors

Aug
30
2008 in Epic Fail, Mono, Bug Reports, C:SI and WTF.

19 Comments

Yup, that’s right, I said it : The Mono rollout blows goats.

The plan, as given by Linden Lab™, was that existing scripts would not need to be recompiled to Mono, and could continue to run as they always have :

Unless scripts are explicitly converted to run on Mono they will continue to run on the existing scripting engine. There is no automated conversion to Mono and all existing scripted content will continue to run on the original scripting engine as before.

That is a ~~great~~good plan, and if it had worked as intended I would have had no problems with it. The problem is, of course, that it doesn’t work. Epic Fail.

Anyone who has tried to duel since the Mono server rollout knows what I’m talking about : The swords can take upwards of 5 seconds to switch state from block to attack or vice versa. Apparently the "run on the original scripting engine" part was left behind in favor of a "run on a (mostly) compatible scripting engine" philosophy.

While trying to come up with an interim fix, I tried compiling the C:SI scripts already in the Taketori to Mono, with mixed results. Some scripts *still* don’t compile (and why the hell would they roll out Mono when everything still doesn’t compile?!?!??!), though most of the ones I tried did compile. The difference was noticeable, so I gave a few Mono-compiled betas out to get some feedback.

From what I can gather by that feedback, I’m not going to get away with such a quick and easy workaround : The scripts still have a very noticeable delay when switching states, and do not perform at all as well as the original scripts used to.

To say I’m upset and discouraged would be an understatement. In fact, a good way to demonstrate how irritated I am without a long litany of curse words is just to quote one of Judo’s comments (emphasis added) :

Judo Jung said:

So I am curious, what is the game plan now given that C:SI is effectively broken? Are the developers going to work on mono, and will we have to wait until the scripts have been rewritten for mono to fight again, and if so, any estimate of how long that might take? Or will there be any attempt to deal with the block latency problem using the existing script? Without C:SI, SL is a bust for me!

Thanks

Yeah.

I’ll re-iterate that Archanox is working on the long-term solution of refactoring the scripts into a real working system under Mono, though of course there’s just now way of knowing how long that can take. I wouldn’t make any bets on it being in the immediate future, though.

Until then, I guess we can somehow try to let Linden Lab™ know of the problem and how it affects close to twenty thousand people in-world, but given their very poor history of responding to such issues and the ever-dwindling avenues of communication I can’t promise that would be any quicker. Maybe all of us together can make enough noise to be heard, who knows?

I will work on trying to figure out how to either beat the existing scripts into some semblance of a workaround, or see what options might be available to get everything working as well as possible.

We will do our best to get this issue resolved as quickly as possible, and we will keep you all informed as we go.

In the meantime, I have suspended sales of the Taketori, as I don’t expect that I’ll have enough time to respond to the anticipated customer messages about buying a broken product

[UPDATE]

This comment was left on the "What in the world is wrong with Second Life" post :

Babbage Linden said:

The only issue I know if which affects scripts running on the original scripting engine in 1.24.3 is that state changes are incorrectly restricted to 1 per frame. We have a fix for this issue which will be likely rolled out next week. If the problem is due to something else, please file a bug on the public JIRA here https://jira.secondlife.com/browse/SVC-1276

Babbage Linden’s latest blog post is : Mono Launch

I’ve been working on a workaround, and I think I have found out something interesting about the cause. I have several beta testers trying out some changes, and I’m working on creating a repro case for Linden Lab that focuses on just the problem at hand so that I can provide them with as much relevant information as possible.

Thank you, Babbage Linden, for taking time to brave these hostile waters and leaving a comment. It is much appreciated!!!

Popularity: 42% [?]

Share This

Off Topic - 1100 Barrel Paintball Gun

Aug
29
2008 in off topic, teh funny and Cool Tech.

0 Comments

What an amazing way to illustrate the different between serial and parallel processing, which is what makes video cards such a powerful general computing platform (for certain classes of problems, at least) in addition to being great at graphics.

I think they did an excellent job on this