Hi Mark, On Mon, Feb 3, 2025 at 8:36 AM Mark Wielaard <m...@klomp.org> wrote: > > I think this subtly changes what is being tested here. If I remember > correctly this tests that gelf_getshdr works correctly even when the > underlying fd is closed and the file is either mmapped or elf_cntl > with ELF_C_FDREAD is called. > > The idea being that if the file is mmapped then everything needed for > reading the ELF file is already in memory. And when elf_cntl is called > with ELF_C_FDREAD then libelf must make sure the same is true. > > You could argue that the test should explicitly call elf_cntl with > ELF_C_FDDONE after the close (fd) though. > > So I don't think close should be called unconditionally. Maybe the > correct solution is to do if (!close_fd) close (fd); after elf_end?
Thanks for the clarification. I think you're right about simply including `if (!close_fd) close (fd)` at the end. The test still passes for me with that change. I've dropped this patch and instead included the `if (!close_fd)...` change in patch 5/5 "tests: Avoid leaking file descriptors". Are you ok with me pushing the series with this change or would you prefer I repost the series? Aaron
