I /think/ I might have broken something - this render of the benchmark
scene doesn't looks quite right... >_<

Post a reply to this message

Attachments:
Download 'benchmark.png' (6 KB)

Preview of image 'benchmark.png'

clipka <ano### [at] anonymousorg> wrote:
> I /think/ I might have broken something - this render of the benchmark
> scene doesn't looks quite right... >_<

heh, well, only way I can get that same thing is by commenting out everything
after the fog statement. If that was the reason it's more like a scene deletion
than a bug!  :)

Post a reply to this message

Am 05.12.2016 um 06:56 schrieb omniverse:
> clipka <ano### [at] anonymousorg> wrote:
>> I /think/ I might have broken something - this render of the benchmark
>> scene doesn't looks quite right... >_<
> 
> heh, well, only way I can get that same thing is by commenting out everything
> after the fog statement. If that was the reason it's more like a scene deletion
> than a bug!  :)

No, I'm currently refactoring the bounding hierarchy code. So I'm quite
sure I've taught POV-Ray to utterly fail to look up any objects in the
bounding box tree. Or the BSP tree if +BM2 is used. Or the simple list
of objects when +BM0 is used.

Which in a sense means I'm on the right track: Until now all the three
modes used completely different code. Now I've made them completely
interchangeable, hiding behind a common interface.

So the fact that I get the same broken picture regardless of the +BMn
setting means that the implementations of the different bounding
hierarchies have been adapted to the new interface consistently; and
that the render engine does indeed consistently use the new common
interface... just not /correctly/ ;)

Post a reply to this message

On 2016-12-05 2:01 AM, clipka wrote:
> Am 05.12.2016 um 06:56 schrieb omniverse:
>> clipka <ano### [at] anonymousorg> wrote:
>>> I /think/ I might have broken something - this render of the benchmark
>>> scene doesn't looks quite right... >_<
>>
>> heh, well, only way I can get that same thing is by commenting out everything
>> after the fog statement. If that was the reason it's more like a scene deletion
>> than a bug!  :)
>
> No, I'm currently refactoring the bounding hierarchy code. So I'm quite
> sure I've taught POV-Ray to utterly fail to look up any objects in the
> bounding box tree. Or the BSP tree if +BM2 is used. Or the simple list
> of objects when +BM0 is used.
>
> Which in a sense means I'm on the right track: Until now all the three
> modes used completely different code. Now I've made them completely
> interchangeable, hiding behind a common interface.
>
> So the fact that I get the same broken picture regardless of the +BMn
> setting means that the implementations of the different bounding
> hierarchies have been adapted to the new interface consistently; and
> that the render engine does indeed consistently use the new common
> interface... just not /correctly/ ;)
>

See - you're making progress :-).  It's sometimes hard to imagine that 
the transition from "inconsistent" to "consistent" sometimes passes 
through "consistent but broken" first and that can be considered a step 
forward.

David Buck

Post a reply to this message

clipka <ano### [at] anonymousorg> wrote:
> Am 05.12.2016 um 06:56 schrieb omniverse:
> > clipka <ano### [at] anonymousorg> wrote:
> >> I /think/ I might have broken something - this render of the benchmark
> >> scene doesn't looks quite right... >_<
> >
> > heh, well, only way I can get that same thing is by commenting out everything
> > after the fog statement. If that was the reason it's more like a scene deletion
> > than a bug!  :)
>
> No, I'm currently refactoring the bounding hierarchy code. So I'm quite
> sure I've taught POV-Ray to utterly fail to look up any objects in the
> bounding box tree. Or the BSP tree if +BM2 is used. Or the simple list
> of objects when +BM0 is used.
>
> Which in a sense means I'm on the right track: Until now all the three
> modes used completely different code. Now I've made them completely
> interchangeable, hiding behind a common interface.
>
> So the fact that I get the same broken picture regardless of the +BMn
> setting means that the implementations of the different bounding
> hierarchies have been adapted to the new interface consistently; and
> that the render engine does indeed consistently use the new common
> interface... just not /correctly/ ;)

Wish I could help but my programming know-how is limited to Basic and Fortran,
both long since forgotten! Otherwise that sure sounds like a missed call out to
a subroutine. LOL
Hopefully something simple like that anyhow.

Bob

Post a reply to this message

Am 05.12.2016 um 13:43 schrieb David Buck:

>> So the fact that I get the same broken picture regardless of the +BMn
>> setting means that the implementations of the different bounding
>> hierarchies have been adapted to the new interface consistently; and
>> that the render engine does indeed consistently use the new common
>> interface... just not /correctly/ ;)
> 
> See - you're making progress :-).  It's sometimes hard to imagine that
> the transition from "inconsistent" to "consistent" sometimes passes
> through "consistent but broken" first and that can be considered a step
> forward.
> 
> David Buck

Thanks -- it's a great consolation to hear that from The Man Himself ;)

(Though now I feel even more ashamed to have broken what you broght to
life ;))

Post a reply to this message

Am 05.12.2016 um 13:47 schrieb omniverse:

> Wish I could help but my programming know-how is limited to Basic and Fortran,
> both long since forgotten! Otherwise that sure sounds like a missed call out to
> a subroutine. LOL
> Hopefully something simple like that anyhow.

Nope.

I give you another hint: This is how it looks with the height_field set
to "no_image" ;)

Post a reply to this message

Attachments:
Download 'benchmark.png' (61 KB)

Preview of image 'benchmark.png'

clipka <ano### [at] anonymousorg> wrote:
> Am 05.12.2016 um 13:47 schrieb omniverse:
>
> I give you another hint: This is how it looks with the height_field set
> to "no_image" ;)

Okay, now that would scare me. Just how convoluted is the source code anyway?
Wait, don't answer that, I probably don't want to know and couldn't understand
anyhow!

For some reason this makes me think of a fun house hall of mirrors at the fair.

Post a reply to this message

omniverse <omn### [at] charternet> wrote:
> clipka <ano### [at] anonymousorg> wrote:
>> Am 05.12.2016 um 13:47 schrieb omniverse:
>> 
>> I give you another hint: This is how it looks with the height_field set
>> to "no_image" ;)
> 
> Okay, now that would scare me. Just how convoluted is the source code anyway?
> Wait, don't answer that, I probably don't want to know and couldn't understand
> anyhow!
> 
> For some reason this makes me think of a fun house hall of mirrors at the fair.
> 
> 
> 

You should see the code map I generated from source!

It eats children, poops in your coffee and molests your dog....when it is
in a GOOD mood. ;-)

Ian

Post a reply to this message

Am 05.12.2016 um 17:24 schrieb omniverse:
> clipka <ano### [at] anonymousorg> wrote:
>> Am 05.12.2016 um 13:47 schrieb omniverse:
>>
>> I give you another hint: This is how it looks with the height_field set
>> to "no_image" ;)
> 
> Okay, now that would scare me. Just how convoluted is the source code anyway?
> Wait, don't answer that, I probably don't want to know and couldn't understand
> anyhow!

To the contrary, it didn't scare me in the least -- it just confirmed my
guess: Turns out I just got the test for all the "no_whatever" flags
backwards.

Post a reply to this message