I see the light!

This snap was taken at dawn:

http://img519.imageshack.us/img519/2907/shinywallsection.jpg

The bright strip of brickwork is reflecting sunlight from the z face of the prim (the house is on an island and there is no other source of light opposite). Please note that I haven't set bumpiness or shiny on that middle prim. I checked each face individually to make sure of this.

The prim does not reflect sunlight at any other time in the SL day.

The only difference I can see between this prim and its neighbours is that it is rotated 90 degrees on the z-axis. The sections of wall either side of it are also z faces but rotated 270 degrees. All copies of this prim on the same wall also reflect light in the same way.

Any suggestions what might be going on here?

It isn't any surface at all that is doing the reflecting. At certain times of day, in Sunset and in Sunrise, the client screws up and for a moment sends light from completely the wrong direction. I see it happen in my beachfront cottage on the West shore of an isolated sim, with no prims West of me. At sunrise, with the sun still way East of me, I suddenly get illumination coming at me from the West! And there's no prims at all West of me for it to be reflecting off of.

Thanks for that, Ceera. What concerns me here is that the effect is selective. I wouldn't mind so much if all the prims in the wall looked the same one way or the other.

I can't think of a good a reason why I had them arranged in this alternate fashion. The walls are very thick - three quarters of a metre - and all the pieces are tapered to make deep windows in the style of an old stone cottage. The reflective pieces are tapered inwards, which is the only essential difference. I guess I should have cut them on consistent faces but I was in a hurry and there you go.

Anyway, I've re-oriented all the prims to face in the same direction. Hopefully they will all react to the sun in the same way. I'll post when I see the result.

ETA: Despite rearranging all the reflective prims in the wall at the same rotation as those that don't reflect light, the issue persists. Apart from vertical height, the only other difference between those that reflect and those that don't is the tapering parameters: the reflective prims are tapered by 0.5 along the x-axis whereas the non-reflective prims are tapered by 0.4 on the same axis and none of the tapered faces is touched by light.

It's an ugly effect but I suppose I'll just have to live with it for the sake of the few minutes it persists.

Perhaps the tapering is affecting the vertex lighting solution?

As an experiment, what happens if you taper all the prims identically, first at .4, and then again at .5?

Thanks for the suggestion, Chosen, but it doesn't have any effect. I altered everything about the prims to match - I even rotated them so that they were more or less identical prims sitting side by side and essentially useless to the build - but the effect persists. It's the damnedest thing.

Ceera's explanation of a rendering issue in the client fits because I see the problem on the northwest and southwest faces as well as the east face, which is the front of the building facing the sunrise of course. Neither face should be catching the sun at all.

In the past I've noticed that lighting direction appears to be a little vague around sunrise but this is the first time I've seen similar prims reacting in such different ways.

What happens if you clone the prims and move them? Do they retain their individual lighting weirdness?

The problem recurs on the cloned prims but at least the effect is seamless.

I suppose I could do just that if I want to keep the deep style of the windows and avoid the neurotic lighting effect. However that would require an extra prim on the inside, sheared to fill the notch between the tapered prims, which is a little extravagant even for a prim wastrel like myself. I guess I'll just have to live with it.

Who's around at dawn anyway?

Hmmm - can you replace the odd prim with a modified version of one of the still unaffected cloned prims?

That's what I did with Chosen's suggestion. I altered the parameters of the unaffected prim to match the 'shiny' prim but it reacted to the lighting in exactly the same (wrong) way.

The lighting discrepancy occurs as soon as I re-adjust the tapering to fit the build.

I can only think that it must be something to do with the rendering of polygons. If I accept that client handling of the lighting is unstable for a few minutes when the sun is near the horizon, it follows that the orientation of faces on the prims will be unpredictable. Only exactly similar prims show the same effect.

Otherwise, everything is exactly square in terms of angle and position: I avoid odd numbers and fractions on the other side of the radix point like the plague.

I guess it's something to mark up as 'just one of those things'. Maybe Viewer 2.0 (or whatever it gets named) will fix it

A couple of hours ago I learned that a visitor to the parcel couldn't see the effect. There is a possibility that she didn't set exactly the right time on the Day Cycle Editor since the problem only lasts for a few minutes and that might be difficult to locate on the sliders. Nevertheless I figure she probably knew what she was doing.

So it seems it may be related to my Windlight settings. It would be interesting to know if Ceera has something similar in her settings that may cause this to happen on the parcel she mentioned above.

Another new development is that a fairly dim light source placed just in front of the wall will accentuate the effect so that prims look much brighter still. However, if I place the light source with its centre on the surface of the wall it cancels out the effect completely across the entire wall.

Apologies for bumping this folks but my heart is broken looking at this awful effect. It turns out it isn't caused by my Windlight settings after all. I don't use Windlight for fancy effects - just enough to soften the high contrast of the default settings. I only got around to properly testing the build with the default Windlight settings today and it actually looks worse.

Furthermore, it seems that any light source will cause the problem: I discovered this earlier when I hung a coach lamp over the door. This snapshot shows the Windlight default setting for midnight and the effect of the light on that wall section:

http://img526.imageshack.us/img526/6310/madlighting.png

The light has a radius of 1 metre, intensity of 1 and fall off is 0.75. The yellow light can also be seen on the prim under the next window in the top left corner of the screenshot. Even though it's much weaker there, the disparity is still very noticeable with the unlit section of wall between the two.

The effect persists with the light cast by the lamp throughout the day - as long as there is sufficient lighting contrast to pick up the colour. I don't plan on keeping the lamp lit throughout the day but it seems I won't be able to get away with any lighting at all on this build if I can't figure out exactly what is causing this.

Any chance you could drop me copies of the offending wall sections so I can fiddle with them when I get a chance? I like puzzles

If the textures are not significant and/or not copyable then untextured is fine. I'd just like to start with known 'bad' prims rather than try to recreate the effect.

Hitting the road shortly for a trip to the mountains but will have some spare time over the weekend.

Cheers
Rime

Same for me This effect sounds like i want to know more about it...
So if you would also drop me a copy in world, i would eagerly do some experiments with it.

Sure. I'll drop copies of the entire wall along with details of orientation and position to both of you.

One other thing occured to me, which is the possibility that something might be rotten in the state of my debug settings although I don't usually play with these unless I know exactly what I'm doing. As with Windlight, I avoid making any tricky or involved changes that might cause my client to act very differently to the default settings but you never know.

So it turns out this has nothing to do with my graphics settings at all. It is a server-side issue.

I made a few more experiments of my own with some friends and it seems that lighting effects on the surface of tapered prims vary, growing darker or brighter depending on the prim's position in relation to the lighting source. Not only that but Pathcut, Slice and Hollow get much brighter as the values are increased. I also made a copy of the wall with all tapers and pathcuts removed and although the visible effect was reduced, it was still very noticeable.

This JIRA from last August seems to be closest to the issue: https://jira.secondlife.com/browse/SVC-4706

While i have no solution or workaround, i now at least understand, what happens. I have recorded a 2 minutes video. I think that "says" more than 1000 words

http://www.vimeo.com/8635612

oof, sure enough... it looks like it's being applied evenly without relation to lighting source and prim relative position (with some modifiers for shape).

does the same result occur if it's unlinked?

Wow... That really shows that bug clearly, Gaia!

Come to think of it, I've seen the same issue in 3D rendering apps like DAZ|Studio. If you stack surfaces, and light them with a common light source, there also there is a tendancy for the light distribution to be calculated from the corner of the face, and not based solely on distance from the origin of the light. So this may well be an OpenGL issue, and not specific to Windlight at all.

I had a similar thought, if the same occurs with them unlinked it would suggest it, but if it doesn't then it'd be relative positions of link parts not being fully reported for lighting. although neither case seems definitive, it'd suggest one or the other.

Hummm. Well, I tried an experiment in DAZ|Studio, to see if this was common to all OpenGL Renders. Nope. Whatever it is, DAZ3D fixed it in their DAZ|Studio app. I can't reproduce that bug any more in that app.

Thanks for that, Gaia. I love the melodramatic soundtrack: it suits my feelings about the issue perfectly.


Linking/unlinking makes no difference.

I tried every alternative I could think of including 'cladding' or 'siding' the bulid with plain, untapered, uncut slabs on the surface but it was just as uneven. Worse still, I thought I'd try hollowing a window in one of the slabs and the effect went apesh!t at the first increment of 0.10 with a weird and very bright prismatic effect extending from the centre out to the corners.

I also tried placing a second lighting source inside the build. It gave the room a checkerboard look with every plane having a flat but very different lighting effect.

I sincerely hope the Lindens do something about it - and soon. Lighting is such an important feature.

They are... watch the end of the video...

http://www.youtube.com/watch?v=swh6gY_dEH0

Can't wait for mesh support or global illumination!

I can't view YouTube, Johan. What's the gist of it?

end of it demos some very nice global lighting effects.