Here are the test cases for the zero_bss fix, as Richard requested. I've included two tests that trigger the bug: - test-zerobss-1.c: minimal case that just validates the loader handles it - test-zerobss-2.c: actually reads from .bss to verify it's properly zeroed
I also tested a third case where the .bss is file-backed in the R_X segment, but that one fails with EINVAL at mmap() time (before reaching zero_bss). That's expected behavior (kernel blocks it) so I didn't include it as a test. I mentioned it in the first commit message though. Thanks, Razvan
