gz-sim: 👩‍🌾 Windows build failures: Unknown CMake command "PROTOBUF_GENERATE_CPP".

Environment

OS Version: Windows
Source or binary build? source, branches off main (may also happen for ign-gazebo5 branches)

Description

Expected behavior: CI passes on Windows
Actual behavior:

CI has been flaky on Windows with this error:

CMake Error at src/msgs/CMakeLists.txt:1 (PROTOBUF_GENERATE_CPP):
  Unknown CMake command "PROTOBUF_GENERATE_CPP".

I’m not sure yet what’s causing the issue. See for example these 2 builds with the same setup, but different results:

- From #961
- That commit has set(protobuf_MODULE_COMPATIBLE TRUE)
- Ran on the windows_mad node
- From #983
- That commit has set(protobuf_MODULE_COMPATIBLE TRUE)
- Ran on the windows_mad node

Steps to reproduce

Run Windows CI a few times for branches off main or ign-gazebo5

https://github.com/osrf/buildfarmer/issues/224

About this issue

Original URL
State: closed
Created 3 years ago
Comments: 21 (21 by maintainers)

Most upvoted comments

@traversaro I updated #1168 with your suggestions and still failed: https://build.osrfoundation.org/job/ign_gazebo-pr-win/3366/

Blast545 on Nov 11, 2021

@traversaro I didn’t consider it. Does it make a difference when the build process reaches https://github.com/protocolbuffers/protobuf/blob/3.13.x/cmake/protobuf-options.cmake#L6 if the variable is defined in the cache?

I may be wrong, but I think that if protobuf_MODULE_COMPATIBLE is in the cache, the option command does nothing (so it does not change the value of protobuf_MODULE_COMPATIBLE) regardless of the value of CMP0077, but I may be wrong.

traversaro on Nov 10, 2021

I’ll take some time to reproduce this error locally consistently before opening a new fix PR though.

Blast545 on Sep 29, 2021