This script calculates the horiz and vert direction based on the angle. I think we are seeing different results from different PCs. This test will allow us to check to see if the math is consistent across GIMP installs, OS, & Hardware.

1) Install the attached test-calc.scm script
2) Open any image and run Script-Fu/Test Calc
3) Post the results displayed on error console back to this thread.

Here are the results of my tests.



Thanks!

Here are my test calc results -

Intel Celeron N282 Intel HD graphics 4GB memory Windows 8.1 64 bit GIMP 2.8.14

Gimp 2.8.14, Ubuntu 14.10
Memory: 2.0 GiB
Processor: Intel Core 2 Duo CPU T5550 @ 1.83GHz
Graphics: Intel GM965/GL960 x86/MMX/SSE2
OS type: 32-bit

Test Calc Warning
Angle(0) horz: 1 vert: 0

Test Calc Warning
Angle(45) horz: 1 vert: 1

Test Calc Warning
Angle(90) horz: 0 vert: 1

Test Calc Warning
Angle(135) horz: 0 vert: 1

Test Calc Warning
Angle(180) horz: 0 vert: 0

Test Calc Warning
Angle(225) horz: 0 vert: 0

Test Calc Warning
Angle(270) horz: 0 vert: 0

Test Calc Warning
Angle(315) horz: 1 vert: 0

Test Calc Warning
Angle(0) horz: 1.0 vert: 0.0

Test Calc Warning
Angle(45) horz: 1.0 vert: 1.0

Test Calc Warning
Angle(90) horz: 0.0 vert: 1.0

Test Calc Warning
Angle(135) horz: -1.0 vert: 1.0

Test Calc Warning
Angle(180) horz: -1.0 vert: 0.0

Test Calc Warning
Angle(225) horz: -1.0 vert: -1.0

Test Calc Warning
Angle(270) horz: -0.0 vert: -1.0

Test Calc Warning
Angle(315) horz: 1.0 vert: -1.0

Below's my result Tux.

Thanks for the tests.

So far, I see Odin is the odd man out. His system does not output the same values as the other results. I'll bet this will be the same as Issabella's results, when she has a chance to post them.

When I released the 3D Extrusion script, I certainly didn't expect that the internal calculations would yield different values across different PCs, but there it is.



On the machines in question, like Odin's, negative values end up being zero, instead of -1.

Anyone have any suggestions as to why this would be the case on certain machines?

Math Co-Processor error? I heard that there might be one on new Intel processors. See link below Tux.

https://randomascii.wordpress.com/2014/ ... intillion/

Looks like the round function isn't doing what's expected of it: http://www.scheme.com/tspl3/objects.html#./objects:s93

And I think the lack of decimal points is indicative of something being different.

If Odin would like to try opening the script-fu console and trying this:

I get the answer -1.0 on GIMP 2.8.14 on my Windows 8.1 machine
I get the answer -1 on GIMP 2.6.7 on my Windows 8.1 machine
I get the answer -1.0 on GIMP 2.8.14 on my Ubuntu 14.04 LTS virtual machine

Kevin

p.s. Script-Fu has a constant: *pi*

And just to throw in a possible reason for the different behaviour. If I add this to your test_calc script:

I can redefine what the round function returns:


So it could be another script redefining round for it's own purpose.

Here my test Calc. Thanks GnuTux

@GnuTux
looks like Issabella and Odin's machines are truncating rather than rounding ..hence the lack of decimal places as well.

Here is one more result, Gimp 2.6.11, Windows 7 home premium, x64.

amd64/quad core/w7/2.8.14

Hello.
Kubuntu Linux 14.04 64bits
intel i7 3770
Gimp 2.8.14

Here's mine a bit late.

Not sure this is a good way to obtain a direction. What happens for small angles (sin or cos <.5)?

You should be using the sign, and if there is no such function in Scheme, you can hack one by dividing the variable by its absolute value (with special case for 0).

But really, nothing wrong with something like

For vpos:



for hpos:

Gimp 2.8.14, Ubuntu 14.10

This script can only produce extrusion at angles which are multiples of 45 degrees - at each step of the extrusion process it simply copies the existing layer to a new layer, translates it by {-1, 0 or +1} pixels in each of the x and y directions and then merges with the existing layer. The fact that the GUI allows you to enter intemediate angles is rather misleading - eg. if you enter 30 degrees it will just round down to 0; if you enter 31 degrees it will round up to 45.
So it may just as well be done with a look-up table anyway!

But that's missing the point vis-a-vis the math problem - the (round x) function should work and it should work the same on all platforms.

Could you (or Oregonian, or Isabella) open up the Script-fu console ("Filters->Script-fu->Console") and report what is displayed if you type "round" by itself (no quotes)?

If the output is:
#<round PROCEDURE 66>
then 'round' has not been redefined by another script. However, if the output is:
#<CLOSURE>
then it has been re-defined.

For the latter case, please type the following command into the console and report what is displayed:
(get-closure-code round)

(This will display the new definition which might prove useful in tracking down what script has re-defined it. At least it will show the faulty definition.)


If the problem is not owing to re-definition of 'round', please open the GIMP Python console ("Filters->Python-Fu->Console") and see if the function behaves there:
round( -0.7)

Oops, the first sentence in my preceding post should have been directed at Odinbc. I hadn't realized that he was quoting paynekj.