I run a llvm profile on my home server and wanted to switch to pypy. Emerge breaks when generating CFFI modules, during the first module pypy_util already. I copied the build to a local overlay and put a die command before the actual generation process. When running the commands manually, it generates without error, in the ebuild it always fails. Reproducible: Always Steps to Reproduce: 1. emerge pypy Actual Results: Breaks with error message ./_pypy_util_cffi_inner.o: file not recognized: file format not recognized Expected Results: Generating all CFFI modules properly
Created attachment 918893 [details] emerge --info
Created attachment 918894 [details] build.log
When you run it manually, CC and so on probably isn't set. You can try sourcing `/var/tmp/portage/dev-lang/pypy-3.11.7.3.18_p1/temp/environment` in the shell and I bet it'll fail, then you can try narrow down what is making it fail.
You were so right, setting me on the correct course to continue looking. Eventually, it turned out my CLFAGS were causing the issue, more specific `-flto=thin` I suggest changing the ebuild to filter out unsupported CFLAGS.
