Custom shadow caster materials

[WilliamLewww]

I added a new SDF parameter ignition:shadow_caster_material_name to specify what shadow caster material should be used to render shadows.

<scene>
  <ignition:shadow_caster_material_name>Gazebo/shadow_caster_ignore_heightmap</ignition:shadow_caster_material_name>
  <background>0.0 0.0 0.0 1</background>
</scene>

Currently, the scene is hard-coded to always use "Gazebo/shadow_caster" for all shadow caster materials.

Adding custom shadow casters could also allow the usage of custom uniform parameters without clumping up "Gazebo/shadow_caster".

Related pull request: WilliamLewww#5

[scpeters]

I managed to remove the input parameter to the service in scpeters@18092a2; please consider merging it in

[WilliamLewww]

The changes work great! I just merged in your branch.

[scpeters]

the CustomShadowCaster tests are failing on macOS:

• https://build.osrfoundation.org/job/gazebo-ci-pr_any-homebrew-amd64/2493/testReport/junit/(root)/CustomShadowCasterTest/DefaultShadowCaster/
• https://build.osrfoundation.org/job/gazebo-ci-pr_any-homebrew-amd64/2493/testReport/junit/(root)/CustomShadowCasterTest/CustomShadowCaster/

I think it's ok to disable these tests if there is not an easy fix

[WilliamLewww]

the CustomShadowCaster tests are failing on macOS:

* https://build.osrfoundation.org/job/gazebo-ci-pr_any-homebrew-amd64/2493/testReport/junit/(root)/CustomShadowCasterTest/DefaultShadowCaster/

* https://build.osrfoundation.org/job/gazebo-ci-pr_any-homebrew-amd64/2493/testReport/junit/(root)/CustomShadowCasterTest/CustomShadowCaster/

I think it's ok to disable these tests if there is not an easy fix

These tests I created use approximating for validation and can be unpredictable.
I think the build not failing is a good test, but I'll fix up the tests for completion :)

[WilliamLewww]

The test is still failing and I'll try to update it again before I just remove it.

[scpeters]

I did notice an issue with PCF on macOS in #3036 (review), but I'm not sure if that's related

[WilliamLewww]

#3036 uses separate shaders so that might not be the issue. I removed the integration test to prevent future error output.

[scpeters]

#3036 uses separate shaders so that might not be the issue. I removed the integration test to prevent future error output.

the test is working on Ubuntu, so I would add it back in but disable it for macOS. For example, here is some cmake logic that skip tests depending on the operating system:

• https://github.com/osrf/gazebo/blob/10969364534f525788001c5409497e5e412af87b/test/integration/CMakeLists.txt#L126-L131
• https://github.com/osrf/gazebo/blob/10969364534f525788001c5409497e5e412af87b/test/performance/CMakeLists.txt#L21

[Tobias-Fischer]

Hi @WilliamLewww @scpeters,

We received a bug report in the conda-forge repo (conda-forge/gazebo-feedstock#90) that Gazebo 11.8 (where this PR first appears) does not start in Windows, with this error:

[Err] [..\gazebo\gui\main.cc:37] Ogre Error:ItemIdentityException: Resource with the name shadow_caster_vp_glsl already exists. in ResourceManager::add at ..\OgreMain\src\OgreResourceManager.cpp (line 158)
[Err] [..\gazebo\server_main.cc:57] Ogre Error:ItemIdentityException: Resource with the name shadow_caster_vp_glsl already exists. in ResourceManager::add at ..\OgreMain\src\OgreResourceManager.cpp (line 158)
[Err] [..\gazebo\transport\Connection.cc:273] The I/O operation has been aborted because of either a thread exit or an application requestValue[995]

I think that this PR is the culprit. Do you have any idea how to fix the issue? I am not sure whether the problem occurs outside of our conda-forge build.

/cc @johnwason @wolfv @traversaro @seanyen

[WilliamLewww]

Hi @WilliamLewww @scpeters,

We received a bug report in the conda-forge repo (conda-forge/gazebo-feedstock#90) that Gazebo 11.8 (where this PR first appears) does not start in Windows, with this error:

[Err] [..\gazebo\gui\main.cc:37] Ogre Error:ItemIdentityException: Resource with the name shadow_caster_vp_glsl already exists. in ResourceManager::add at ..\OgreMain\src\OgreResourceManager.cpp (line 158)
[Err] [..\gazebo\server_main.cc:57] Ogre Error:ItemIdentityException: Resource with the name shadow_caster_vp_glsl already exists. in ResourceManager::add at ..\OgreMain\src\OgreResourceManager.cpp (line 158)
[Err] [..\gazebo\transport\Connection.cc:273] The I/O operation has been aborted because of either a thread exit or an application requestValue[995]

I think that this PR is the culprit. Do you have any idea how to fix the issue? I am not sure whether the problem occurs outside of our conda-forge build.

/cc @johnwason @wolfv @traversaro @seanyen

I think I see the issue where shadow_caster_vp_glsl is defined both in:

• gazebo/media/materials/scripts/shadow_caster.program
• gazebo/media/materials/scripts/shadow_caster_ignore_heightmap.program

A temporary fix could be to:

1. Delete the following lines
   https://github.com/osrf/gazebo/blob/b502f17729302fa0da8e2a00ffbd69112bb7e6aa/media/materials/scripts/shadow_caster_ignore_heightmap.program#L1-L10

If that doesn't work then
2. Change
https://github.com/osrf/gazebo/blob/b502f17729302fa0da8e2a00ffbd69112bb7e6aa/media/materials/scripts/shadow_caster_ignore_heightmap.program#L32
to vertex_program_ref DeferredShading/vs

I'm currently checking out the issue and rebuilding my workspace, but I'll send a follow up if the solution works on my end.

[Tobias-Fischer]

Thanks for looking into it so quickly @WilliamLewww. It would be great if you let us know which of your two suggestions ends up being the better one, and we can patch it on conda-forge until a new release comes out.

[WilliamLewww]

Thanks for looking into it so quickly @WilliamLewww. It would be great if you let us know which of your two suggestions ends up being the better one, and we can patch it on conda-forge until a new release comes out.

No worries! I was able to fix the issue on my end with the following changes: WilliamLewww@8d50434.

I will be opening a pull request soon to add the changes. Let me know if this fixes the issue you ran into.