Maniphest T48790

Wrong render with HDRi when there Sun lamp with strength=0
Closed, Resolved
Actions

Assigned To
Brecht Van Lommel (brecht)
Authored By
YAFU (YAFU)
Jul 5 2016, 5:22 AM
Tags
  • BF Blender
  • BF Blender: Regressions
  • Cycles
Subscribers
Bastien Montagne (mont29)
Brecht Van Lommel (brecht)
Campbell Barton (campbellbarton)
Sergey Sharybin (sergey)
Thomas Dinges (dingto)
YAFU (YAFU)

Description

System Information
Kubuntu Linux 14.04, GTX 960

Blender Version
Broken: from master, 2.77.1 39ae324
Worked: 2.77a

Short description of error
Wrong render with HDRi when there Sun lamp in the scene with strength=0

Exact steps for others to reproduce the error
Scene:

hdri_sun0.blend597 KBDownload

HDRi: https://hdrihaven.com/hdri.php?hdri=sisulu&dl=yes
Open the scene. Load the HDRi (it is not packed). Render.

Revisions and Commits

rC Cycles
rC1f49c7bc43ee Fix T48790: Cycles render bug with zero strength lights
rB Blender
rBb645e7081d2b Fix T48790: Cycles render bug with zero strength lights.

Related Objects

Mentioned In
D12610: Fixes T91632 by stopping the sample correlation between dimensions which was causing rendering artifacts on simple scenes.
rBL61704: Cycles tests: Add test for T48790
Mentioned Here
rB42aec3b355b7: Cycles: nodify shader nodes
rBb27322e71ee4: Curve: utility to evaluate entire curve
rB39ae32491817: Cycles: remove extended precision hacks, no longer needed with SSE2 requirement.

Event Timeline

YAFU (YAFU) created this task.Jul 5 2016, 5:22 AM
YAFU (YAFU) raised the priority of this task from to 90.
YAFU (YAFU) updated the task description.
YAFU (YAFU) added a project: BF Blender.
YAFU (YAFU) edited a custom field.
YAFU (YAFU) added a subscriber: YAFU (YAFU).
YAFU (YAFU) added a comment.Jul 5 2016, 5:34 AM

Sorry, this happens with any type of lamp with strength=0

Campbell Barton (campbellbarton) added a subscriber: Campbell Barton (campbellbarton).Jul 5 2016, 6:00 AM

Renders black when the strength is 0.0, both for 2.77a and master (9269574089a742130f02c0a1184a19d94f0e665d)

YAFU (YAFU) added a comment.Jul 5 2016, 1:19 PM

Hi.
On my machine the render looks good with 2.77a (abf6f08):

Sergey Sharybin (sergey) lowered the priority of this task from 90 to 30.Jul 5 2016, 5:38 PM
Sergey Sharybin (sergey) added a subscriber: Sergey Sharybin (sergey).

The file has Strength of 1.0. Setting it to 0.0 makes everything black (as expected).

  • What are the _exact_ steps reproducing the issue?
  • What is _exact_ issue anyway?
  • Does it happen with CPU?
  • Does it happen with latest builds from builder.blender.org?
YAFU (YAFU) added a comment.EditedJul 5 2016, 6:02 PM

Perhaps I have not been clear here. Apparently you refer to World (HDRi environment texture) strength. You do not touch that value in the scene, you leave that value to 1. With "strength=0" I am referring to the lamp selected by default when you open the scene. You just open the scene and render in Belnder from master (Do not forget to load the HDRi). Did you not get a resulting image like the first image I've shared?. I think it should look like the second image I obtained in Blender 2.77a.
I know, there is no logical reason to put there a lamp with strenght=0. But anyway I do not see any reason to get a resulting image like the obtained from master (first image).
EDIT:
Regarding whether this occurs in CPU/GPU, master/buildbot:

-Buildbot b27322e :
GPU: That diagonal division is very noticeable
CPU: That diagonal division is a bit less noticeable than GPU, but still noticeable.

-Compiled from Master 39ae324 :
GPU: That diagonal division is very noticeable
CPU: That diagonal division barely noticeable, mainly it is more noticeable when it begins to render the first samples.

Bastien Montagne (mont29) raised the priority of this task from 30 to Normal.Jul 15 2016, 5:06 PM
Bastien Montagne (mont29) added subscribers: Thomas Dinges (dingto), Bastien Montagne (mont29).

Can’t confirm this with CPU rendering indeed, don’t have GPU one here… But imho you are stressing system out of expected 'domain', what’s the point in testing something with zero-strength lamp?

Anyway, @Sergey Sharybin (sergey) or @Thomas Dinges (dingto), can you reproduce the issue now with specified steps?

YAFU (YAFU) added a comment.Jul 15 2016, 6:16 PM

Hello Bastien!

In T48790#381753, @Bastien Montagne (mont29) wrote:

what’s the point in testing something with zero-strength lamp?

Oh no. No point at all.
I just was testing another scene where I was working, and I wanted to see the render without the influence of the Sun lamp. As I usually find the outliner a bit complicated, I just decided to put Sun strength=0, then I found the problem . Then I discovered that this did not occur in previous versions of Blender, and then build a small scene and report just in case.
I do not know, maybe in a very large and complex scene a little experienced artist (like me) could be left forgotten there a lamp with strength=0. But I guess it will not be very common cases.

By the way, the problem is still present in my machine and last buildbot.

Sergey Sharybin (sergey) claimed this task.Jul 27 2016, 11:35 AM

@YAFU (YAFU), with the detailed explanation now i see what you mean :) Indeed see the bug here. Investigating...

Sergey Sharybin (sergey) reassigned this task from Sergey Sharybin (sergey) to Brecht Van Lommel (brecht).Jul 27 2016, 2:56 PM
Sergey Sharybin (sergey) triaged this task as 50 priority.
Sergey Sharybin (sergey) added projects: BF Blender: Regressions, Cycles.
Sergey Sharybin (sergey) added a subscriber: Brecht Van Lommel (brecht).

@Brecht Van Lommel (brecht), apparently caused by rB42aec3b. Setting strength is required to make it so you're leaving with lamp shader with just output node. You can also simply remove any connection from lamp shader outputs to see the same issue.

Brecht Van Lommel (brecht) added a comment.Jul 27 2016, 11:13 PM

The issue was in earlier code to disable lights, but better constant folding made it apparent.

Brecht Van Lommel (brecht) changed the task status from Unknown Status to Resolved by committing rBb645e7081d2b: Fix T48790: Cycles render bug with zero strength lights..Jul 27 2016, 11:13 PM
Brecht Van Lommel (brecht) added a commit: rBb645e7081d2b: Fix T48790: Cycles render bug with zero strength lights..
Sergey Sharybin (sergey) added a comment.Jul 28 2016, 9:22 AM

@Brecht Van Lommel (brecht), interesting. Because i was unable to reproduce issue with a revision previous to node refactor. Could have been some red-herring or just some unrelated changes made it visible. Thanks for the fix!

YAFU (YAFU) added a comment.Jul 28 2016, 4:57 PM

Working well now.
Thanks.

Sergey Sharybin (sergey) mentioned this in rBL61704: Cycles tests: Add test for T48790.Aug 3 2016, 4:10 PM
Sergey Sharybin (sergey) added a commit: rC1f49c7bc43ee: Fix T48790: Cycles render bug with zero strength lights.Aug 18 2016, 12:53 PM
William Leeson (leesonw) mentioned this in D12610: Fixes T91632 by stopping the sample correlation between dimensions which was causing rendering artifacts on simple scenes..Sep 24 2021, 9:34 AM