Build failure with ogre 1.9.0 compiled with the OGRE_CONFIG_DOUBLE enabled

[heeplr]

Environment

• OS Version: gentoo
• Source or binary build?
  source: 8.1.1 tag

build.log

Description

• Expected behavior: successful compile
• Actual behavior: compile error
• error: conflicting return type specified for ‘virtual float gz::rendering::v8::OgreMovableText::getBoundingRadius()
• error: no matching function for call to ‘max(float&, Ogre::Real)’

Steps to reproduce

1. compile from source with gcc 13.2.1

Output

[traversaro]

Can you provide the list of dependencies (I am not familiar with gentoo, but I guess there is some way of listing the dependencies). Thanks!

[traversaro]

Also a more complete log error (for example indicating also where the error appeared) could be useful to provide hints on what is going wrong.

[heeplr]

Please see the attached/embedded build.log in the "Details" collapsible.

Dependencies should also be listed in the cmake output. But iirc nothing non-gz specific besides ogre. (I'm currently afk but can provide more details later)

[traversaro]

Ah cool thanks, sorry I missed that. However, I do not see any dependency version listed there. From the some flags, it seems that you are trying to use some ogre version >= 1.12 ? As far as I know, that is not tested and based on my past experience I doubt it works out of the box. For reference, I tried to get ogre 1.12 to work in conda-forge packaging (see conda-forge/libignition-gazebo-feedstock#6 (comment)) and I was never be able to get it to work, so we are pinning ogre to 1.10 to get it to work. Anyhow, most of the focus now in gz-rendering is ogre-next, have you considered packaging and using that?

[heeplr]

Sorry for the confusion, I was in a hurry and didn't fill the form thoroughly.

I'm building againtst ogre 1.9.0, is that too low?

Here's the depgraph:

 * dependency graph for sci-libs/gz-rendering-8.1.1
 `--  sci-libs/gz-rendering-8.1.1  ~amd64 
   `--  media-libs/freeimage-3.18.0-r11  (media-libs/freeimage) amd64 
   `--  dev-games/ogre-1.9.0-r3  (dev-games/ogre) amd64 
   `--  dev-build/gz-cmake-3.5.3  (>=dev-build/gz-cmake-3.5.3) ~amd64 
   `--  sci-libs/gz-math-7.4.0  (>=sci-libs/gz-math-7.4.0) ~amd64 
   `--  sci-libs/gz-common-5.6.0  (>=sci-libs/gz-common-5.6.0) ~amd64 
   `--  sci-libs/gz-plugin-2.0.3  (>=sci-libs/gz-plugin-2.0.3) ~amd64 
   `--  sci-libs/gz-utils-2.2.0  (>=sci-libs/gz-utils-2.2.0) ~amd64 
   `--  app-alternatives/ninja-1  (app-alternatives/ninja) amd64 
   `--  dev-build/cmake-3.27.9  (>=dev-build/cmake-3.20.5) amd64 

I'll try to build ogre-next v2.3.3 and see if it gz-rendering compiles against that.

[traversaro]

Sorry, I was confused by the OGRE_VERSION_LT_1_12_0 macros and read them as OGRE_VERSION_GT_1_12_0. Indeed ogre 1.9.0 is the version used in Ubuntu packaging, so I am not sure what is going on. However, without ogre-next most of the gz-sim example will fail as they default to use ogre-next as rendered, so packaging also ogre-next is for sure a good choice.

[traversaro]

Reading a bit better the error, it seems that the problem is that your ogre build defines Ogre::Real as double, while gz-rendering code base silently assumes that it is defined as float (as I guess is defined in debian, conda-forge and other packaging system builds of ogre). Probably your build of ogre 1.9 is compiled with the OGRE_CONFIG_DOUBLE CMake option enabled (see https://github.com/OGRECave/ogre/blob/v1-9-0/CMakeLists.txt#L393), instead of the default that is off? I guess in theory you could remove this assumptions in gz-rendering with a lot of explicit casts added, but I am not sure how easy it would be.

[traversaro]

(If you could modify the issue title in something more descriptive like "Build failure with ogre 1.9.0 compiled with the OGRE_CONFIG_DOUBLE enabled" it would be great, thanks!

[heeplr]

Ok, I disabled OGRE_CONFIG_DOUBLE (and the equivalent in bullet) and now gz-rendering 8.1.1 builds fine with ogre 1.9.0.

Thanks for your help.