[question] How to detect we're running on c3i?

[madebr]

We're having a lot of problems with building boost packages on c3i.
See e.g. #5420

Often the boost build fails because the test_package is built with a too high parallelism.
The logs contain the line:

MSBUILD : error MSB4166: Child node "6" exited prematurely. Shutting down. Diagnostic information may be found in files in "C:\J\w\BuildSingleReference@7\10837\d953e04c-152c-463f-8164-f2a7e5bf2b0f\" and will be named MSBuild_*.failure.txt. This location can be changed by setting the MSBUILDDEBUGPATH environment variable to a different directory.
  locale.cpp

I think it would be useful to detect we're running on c3i/ci and if so, lower the level of parallelism.
e.g.

cmake = CMake(self)
cmake.configure()
if tools.get_env("CI_C3I"):
    cmake.parallel = False
cmake.build()

What do you think?

[jgsogo]

I would vote to hide this in the config.yml file instead of modifying the recipe.

What about...?

config.yml

versions:
    1.69.0:
      folder: all
    1.70.0:
      folder: all
config:
  env:
     CPU_COUNT: 1  # Will translate to CONAN_CPU_COUNT

or

versions:
    1.69.0:
      folder: all
      env:
        CPU_COUNT: 1  # Will translate to CONAN_CPU_COUNT
    1.70.0:
      folder: all 
      env:
        CPU_COUNT: 1  # Will translate to CONAN_CPU_COUNT

• It will require to develop this feature in the c3i library (not so complicated)
• Probably we need to filter variables that are available and values, just in case.

ping @danimtb

[uilianries]

Why not default Conan env vars, like CONAN_CPU_COUNT? So we don't need to parse, only forward. Of course, we would need a filter to allow only specific env vars, otherwise people can do a mess in CCI.

[madebr]

Building boost in parallel is not causing problems, only the test_package.
So it would be nice to lower the parallelism only there.

[jgsogo]

@uilianries , yes, before thinking about filtering, I thought about pretending always CONAN_ so people cannot set random environment variables. I will update my message so it is not misleading.

[jgsogo]

@madebr , if it is only for the test_package, why not avoiding parallelism always? It shouldn't take much more to build.

[madebr]

That is what I've done in #5420.
Since then, the builds are failing much less.
I was wondering whether that was the correct approach.

[uilianries]

Closing this issue due the new redesigned ConanCenterIndex CI service.

Please, read #25772 for further information.