What is the value of RbConfig::CONFIG['host_os'] and RUBY_PLATFORM ?

Hi @larskanis,

Thanks for your inquiry! Here are the exact values we verified on the RISC-V SG2044 server:

• RbConfig::CONFIG['host_os'] = linux
• RUBY_PLATFORM = riscv64-linux-gnu

These values are confirmed through multiple validations on the SG2044 hardware, and the added RISC-V detection logic works perfectly with this configuration.

Oh, I meant RbConfig::CONFIG['host_cpu'] .

I just wonder why the new case introduced in your patch is necessary at all. FFI::Platform::ARCH should already be riscv64 with and without this pull request. The default branch should be enough.

Hi @larskanis,
The value of RbConfig::CONFIG['host_cpu'] is riscv64. Although the else branch sets ARCH to riscv64, subsequent library loading logic (such as JRubyUtilLibrary.load_ext) may only recognize explicitly listed architectures, resulting in throwing Unsupported platform: unknown-linux. Adding an explicit when /riscv64|riscv/ branch ensures that the subsequent code correctly identifies and handles the RISC-V architecture. We have verified this on SG2044; without the explicit branch, platform recognition errors occur, and adding it resolves the issue.

Adding an explicit when /riscv64|riscv/ branch ensures that the subsequent code correctly identifies and handles the RISC-V architecture. We have verified this on SG2044; without the explicit branch, platform recognition errors occur, and adding it resolves the issue.

That doesn't make sense. Where is the difference between the case branch and the default branch?

I also worry about the second matcher /riscv/ because it will recognize a riscv32 as riscv64.