Le 24/02/2016 16:51, clipka a écrit :
> Am 24.02.2016 um 16:02 schrieb ThH:
>> Am 24.02.2016 um 15:55 schrieb Jaime Vives Piqueres:
>>
>>>    Happens here too on 3.7.1-alpha.8454683, AMD FM-6300, Ubuntu 14.04
.3
>>> LTS.
>>>
>>> -- 
>>> jaime
>>
>> Thanks for checking and confirming Jaime :)
>>
>> So it's not just me ;)
> 
> Okay, here's the deal: I can't reproduce it, so my only chance to
> identify the root cause is to have a close look at the code, but for
> that to make any sense I need a clear picture of exactly where to look.

> To this end, I need you (that's "you" as in "y'all") to:
> 
> - figure out the simplest possible scene that exhibits the problem;
> 
> - identify exactly which commit (in the master branch) broke the scene;
 and
> 
> - do plenty of systematic toying around with the scene to figure out th
e
> exact circumstances under which the error shows. (For instance, why
> isn't the green beam affected? Is it because it is not reflected? Then
> why aren't the red and blue beams affected right after the first
> reflection? Does max_trace_level have anything to do with it? Why does
> only the bottom of the red sub-beams seem affected? Why the top and
> center of the blue sub-beams? etc. -- Be inquisitive and creative!)
> 

The problem might not be only in the code, I tested several heads and
compilers

/=================+=====+==
=====+=======+============
====\
| head \ compiler | gnu | clang | intel | gnu with debug |
+-----------------+-----+-------+-------+----------------+
| 3.7.0 stable    |  ok |  ok   |  ok   |  ok            |
+-----------------+-----+-------+-------+----------------+
| 3.7.1 master    | BUG |  ok   |  ok   |  BUG           |
+-----------------+-----+-------+-------+----------------+
| my master (hg)  | BUG |  ok   |  ok   |  BUG           |
+-----------------+-----+-------+-------+----------------+
| hgpovray        | ok  |  ok   |  ok   |  ok            |
\-----------------+-----+-------+-------+----------------/

notice, the intensity of green on stable branch seems higher (might be
known... gamma ?)

the difference between hgpovray and my master is:
* hgpovray is late on the refactoring of the code
* I have all my additions in hgpovray

the difference between my master and official master:
* I'm about a year late on synchronisation with refactoring

last pull from master to my master was April-June 2015, so the "bug" was
already there then.

The start/fork of hgpovray is 8 june 2014, so code before that date was
still fine.

The OP's picture was in white bug, mine is yellow.
(Ubuntu 64 bits 15.10)

Just one year of patches to check... it might be tied to the size of a
structure, and a particularity of Gnu compiler (gcc/g++ 5.2.1 20151010)
sensible to that... or a change of initialisation...

Post a reply to this message

Attachments:
Download 'gmpovray.png' (239 KB)

Preview of image 'gmpovray.png'