On Mon, May 12, 2025 at 05:17:35PM +0000, Lloyd wrote:
> I gave this some additional thought. Something is not setting an error
> condition when being able to write past end-of-disk when the FS is full.
>
> That means it isn't necessarily reorder_kernel at all, that is the
> symptom, maybe not the cause. The cause could be somewhere deeper
> (linker, filesystem, etc.)
I think the main bug is in the ctfstrip wrapper script of ctfconv. It
doesn't seem to propagate failure to the caller. If that's correct,
then I think the ${SYSTEM_LD_HEAD} and ${SYSTEM_LD_TAIL} targets can't
fail, which in turn means that newbsd can create a garbage kernel (and
that won't fail either since mv -f is used).
Then reorder_kernel continues on its merry way and gives you a broken
boot kernel.
