[Opensim-dev] [libsl-dev] T pose issue

[Melanie]

Hi,

this is different from the reported bug, which was that avatars log
in in T-Pose. This also happens on the LL grid.

However, avatars already in-sim should be correctly displayed to new
joiners.

This is a long-standing issue. It is caused by the current animation
pack not being sent to new observers. On the LL grid, when an avatar
first comes into view, the anim pack is sent to the new observer.

We are, however, currently not tracking "seen" root agents per
client, as we should, so we don't have the data to decide when to
send the anim pack. Because of that, we send it only when it changes.

This issue is distinct from the bug I closed. Please do NOT reopen
that bug, as it doesn't apply to what you're seeing. This is a
different behavior and I'm not aware of an exsting Mantis, although
I've been aware of the issue foerever.

Melanie


Paul Fishwick wrote:
> There is an issue we have when using libomv and the OpenSim
> server, and I am not sure where the problem lies. The issue is that
> our bots log into OpenSim in a T position, and they do not log in
> in a T position on the LL grid. The same libomv software is used
> in both instances. What is the reason for the difference? I had thought
> that it was, perhaps, that the LL server was keeping track of the
> current animation state; however, one of the OpenSim devs says that
> is not so.
> 
> With OpenSim, the problem is that whenever a new human-driven
> client logs into OpenSim, the libomv-driven bot appears as a T pose.
> One must continually change the animation state of the bot to get the
> bot out of the T pose for arbitary human clients who log in.
> We have rigged a few OpenSim bots using repeated animations;
> however, for the bots that are in a seating position on a bus,
> the problem doesn't appear to be fixable--having a bot reseat itself every
> second or so is awkward and would appear unnatural. Using an eye
> animation doesn't do the trick of reseating the bot either.
> 
> On the LL grid, we have many seated bots and LL knows to reseat these
> bots for each human client without animation override.
> -p
>