There were problems with the Tuesday roll out. Some noticed the problems and others of us didn’t. For those that ran into problems it was an annoyance. Things just weren’t working correctly.
Oskar Linden explained what happened:
Things were going great this week ‘til Tuesday then…
So, this is what happened. Our release system is a big machine. It’s operated by hand but it’s still a machine. [I’m visualizing a big Steampunk thing huffing and puffing with levers and stuff.] You put code in one end and it gets released at the other end
What happened was the wrong code was put into the machine so the machine just kept happily shipping the wrong bits. The bits it was shipping were still in QA and not ready for release. It was a mistake. We noticed it after half the grid was deployed.
The code contained some changes to the server interest list and that made some things invisible. So it would look like they wouldn’t rez. It was there. You could see the prim [limit] count decrement. But the server wouldn’t tell the viewer about them.
As soon as I noticed I kinda freaked out. So we went into full emergency mode. [Now I am imagining alarms and sirens going off amid red flashing lights and people running every which way to pull levers and turn wheels while doors are closed and sealed.]We rolled a few regions to the correct version and made sure there weren’t any issues with that then made sure that previously invisible stuff was now visible and then we called it good and restarted the whole roll then the grid settled down and all was good. THE END.
The release candidates destined for the RC channels failed to pass QA this past week (think ADITI testing failed). So, there was no Wednesday roll of new packages to the RC channels. The problem was not that bad and Oskar thinks they might have been able to fix it in time for Wednesday’s roll out. But, because of Tuesday’s hectic day they decided to skip the roll and take more time with the fix.
Oskar says the bug that stopped the RC roll was: when you rez an object, scripts in child prims won’t load.
This means the whole grid is running the same version of server software from Tuesday.
Thursday Oskar though we would see 1 or possibly 2 new RC’s hit the RC channels next week. Simon was saying the same on Friday. Saying there would be no new roll to the main channel as the RC channels are empty. Right. But, two new RC’s will come to the RC channels Wednesday in week 39. Same.
One RC is a maintenance release that includes Baker Linden’s Group Edit code.
The other has some under-the-hood improvements that will help regions run better on new hardware. Nothing dramatic… I suspect this is something to do with the new computer hardware being installed at the Lab. Simon Linden said, “I’ve been working on some server internals that should help prevent them from getting into that zombie state when they are overloaded.”
Mæstro Linden says there is a bug that is about renaming items in an object’s inventory. If 2 items have the same name, because you just renamed one to match the other’s name, you’d lose an item. But classically in that case, we [the SL system] would rename the 2nd one to “item 1” if the other one was called “item”.
BUG-126 – Renaming an item in an objects contents to a name of an already existing item in contents deletes the 1st item.
Andrew was asked if the release of CreatorVerse and Patterns was going to take people away from working on SL. Andrew’s answer: “Not really. We still have the largest percentage of our company working on SL. These products have no real effect on SL.” I assume he means by ‘no effect’ that it does not take people off the SL projects.