Building by the numbers, a word of warning...

If you take a prim and angle it at all, the objects tab might look like they're exact integers (45 degrees, for instance). however, I've reason to believe that if you rotate the objects using the "rotational axis doohickey" instead of punching in the numbers yourself, they're much more exact, and hence nearly impossible to build with.

We're talking .0001 of a meter off, but when you're copying something and watching the x positions go from 10 to 20.045 for no reason, something's up

I suspect another floating point bug.

LF

A related caveat...

If you rez a linked object, then rotate it arbitrarily to something other than +/- 90 degrees or 180 degrees, then unlink and relink, you're practically guaranteed to introduce errors into the object.

They may not be visible right off, but they will accumulate.

The reason is because with a rotation, you're doing a polar conversion on a grid coordinate system.

For example, let's say you have two linked prims. The child prim is <1.0,0.0,0.0> offset from the parent.

If you rotate this linked set 45 degrees, child prim's absolute offset from parent will then be <sin(45), cos(45), 0>. If you unlink and relink the two, it probalby gets saved as <0.70710, 0.70710, 0>. If you then finally rotate the relinked set back back 45 degrees, your child prim will now be at something like <0.99999,0.00001,0.0> offset from the parent due to the lost precision in rounding.

The moral of the story is to always reset your rotation to 0 or 90 degree increments before unlinking to edit, to keep the evil sine and cosine gremlins away =)

thank you kex. i won't have that problem again.

Don't you just love floating point arithmetic precision?

I personaly would love it if they would add great precision with the edit box. With big objects 1 degree is not enough you need half degrees.

With great precision comes great responsibility.





I'm sorry... I just ate a bowl of Coco Pebbles and the sugar rush is going to my fingers.

As I said in another thread earlier today, I'm all for more accuracy in the object manipulation fields. ESPECIALLY if that precision is already in the data structure for the prim!

You can enter numbers to any precision you want. The edit box will round them for display purposes, but internally, they'll still be what you specified.

Carnildo, personal experience would lead me to doubt that. Can you cite a source?

The only source I can cite is personal experience in the building of skyboats.

Slight correction: I'm certain that half-degrees work, and fairly certain that quarter-degrees work. I didn't need any higher precision, so I haven't tested beyond that.

It would also be nice if the position text boxes could display all six digits, instead of cropping the last 1 or 2. Maybe this is display-dependant, but it's *very* annoying to have to select the text and scroll right just to see the last digit.

Example... If the X position is 234.456, the Edit window shows:

X: 234.45

The last digit is cut off because the text box is too small.

Click on the edit box and hit the right arrow key a few times. The text will scroll over and show that last digit.

Well yeah, but don't you think the text box should show all of the digits? They just need to widen all of the text boxes.

100% agree with Cubey about the text box size. What is even more annoying is that I often have problems selecting the last digit. Sometimes I can't click and drag to highlight the last digit. Instead I have to use the arrow key to the right in order to deletethe digit and then type what I want - Very annoying.

Oh, I don't think that 1/2 degrees work, but I will try again.

Thanks for the info on the inacurracy of linked/unlinked/rotated objects. I will be more careful.

I can tell a diffrence between .1 deg diffrences, and is very obvious..(zoomed in on large prims).. But you need to remember they are rounded for display. and to not forget if the value is about .5 or more that it got rounded up for display, so that if you need .7 you remember to take off the 1 deg it showed.. but yeah, am certain .1 deg works from experience.