[R-package] Fix R package cmake build command

[qinwf]

No description provided.

[jameslamb]

Thanks for your interest in LightGBM.

Please update this with a description helping us understand how it improves the project. What problem does it fix? For example, can you share a specific command you can and an error you saw, which this fixes?

[qinwf]

I got an error when building R packages with OpenCL on Linux.

 add in command-line arguments
# NOTE: build_r.R replaces the line below
command_line_args <- c("-DOpenCL_LIBRARY='/usr/local/cuda-10.2/targets/x86_64-linux/lib/libOpenCL.so.1.1'', '-DOpenCL_INCLUDE_DIR='/usr/local/cuda-10.2/targets/x86_64-linux/include'")

The generated install.libs.R is wrong.

[jameslamb]

Can you please share the exact command(s) you're using to run build the R package? So I can try to reproduce the error you saw, to be sure this is the right fix.

[qinwf]

Rscript build_r.R --use-gpu --opencl-library=/usr/local/cuda/lib64/libOpenCL.so --opencl-include-dir=/usr/local/cuda/include

[jameslamb]

Thanks for that! When I run that command on master, like this:

Rscript build_r.R \
    --use-gpu \
    --opencl-library=/usr/local/cuda/lib64/libOpenCL.so \
    --opencl-include-dir=/usr/local/cuda/include

I see the error that I suspect you ran into

Error in source("install.libs.R", local = local.env) :
install.libs.R:143:77: unexpected string constant
142: # NOTE: build_r.R replaces the line below
143: command_line_args <- c('-DOpenCL_LIBRARY='/usr/local/cuda/lib64/libOpenCL.so''

And can confirm that this change fixes it on Linux.

Thanks for that! In the future, when you open pull requests here please provide enough information for us to be able to understand exactly how the change improves the project. And the text of error messages, like I've provided above, so others facing the same issue can find this PR from search.

However.... the proposed change exactly reverts #5607, which was introduced to fix similar quoting issues on Windows, reported in #5135. So if we merge this as-is, it'd again break support for building a GPU-enabled version of the R package on Windows.

Do you have access to a Windows environment to test running the following?

Rscript build_r.R \
    --use-gpu \
    --boost-root="C:\Users\James\repos\LightGBM\external_libs\compute" \
    --no-build-vignettes \
    --use-msys2

If not, then we'll have to close this PR and I will try to put up a fix soon that makes this work on both Unix-like operating system AND Windows.

Sorry it's such a manual process... we don't currently have CI jobs covering the GPU-enabled build of the R package (#3780).

[jameslamb]

Thanks for reporting this issue. I'll interpret your closing this without any comment as you saying "no, this is more complex than I expected and I don't want to help fix it".

I've opened #5960 to document the issue. You can subscribe there to be notified if someone picks it up and submits a fix.

[github-actions]

This pull request has been automatically locked since there has not been any recent activity since it was closed. To start a new related discussion, open a new issue at https://github.com/microsoft/LightGBM/issues including a reference to this.