On Fri, Apr 04, 2025 at 03:33:34PM +0200, Bruno Haible wrote: > Hi Eric, > > Eric Blake wrote: > > While trying to build m4.git with clang on a Fedora 40 machine, and > > with gnulib updated to the latest git commit, I'm hitting the > > following during 'make check' (building with gcc worked): > > The same build configuration (clang on Fedora 40) works for me, > except for a "test-driver: invalid option: '--collect-skipped-logs'" > error at a later point.
Sounds like a separate issue; I'll see if I can reproduce and fix. > > > Is this a case of a missing LDFLAGS in m4's tests/Makefile.am (which > > right now only includes gnulib.mk) or top-level Makefile.am? > > No, rpl_gettimeofday is not meant to come from an external library. > > Here's what I have: > > $ nm tests/time.o > U abort > U gettimeofday > 0000000000000000 T rpl_time > > I would guess that you have some left-over files in your source or > build directory, and would suggest to do > $ make distclean > $ rm -f lib/sys/time.h tests/sys/time.h > $ rm -f `find . -type l` > in both the source and the build directory, before re-running 'bootstrap'. Trying that now, and the first thing I notice is that the 'bootstrap' in use by m4 (Gary's fork, not gnulib's) rewrites the symlinks into gnulib/ to absolute paths rather than relative, which is wrong. Separate bug to fix later... At any rate, it does look like stale files; nuking the world and doing a clean git checkout rather than trying to reuse an existing directory hd no problems. Thanks for the sanity check! -- Eric Blake, Principal Software Engineer Red Hat, Inc. Virtualization: qemu.org | libguestfs.org
