Use a custom tool to detect sparseness.

I honestly hadn't noticed that the tzcode update removed tzsetwall(3).

Add missing md5 part.

Remove comment about the 80386.

Hook up the SHA-512/224 tests.

In D38372#872487, @cperciva wrote:

You're going to be adding test cases for this in libmd, right?

Removed useless .Tn macro from man pages, deliberately not bumping Dd since neither the text nor the appearance has changed.

Note that I intended to add assertions to be64enc_vect() and be64dec_vect(), but since the same code is used both in the kernel and in user space, it would have been needlessly complicated.

Additional cleanup.

In D38291#871933, @ehem_freebsd_m5p.com wrote:

The copyright may be on the weak side for this, but this violates my rights.

Order tests lexicographically.

Abandoning in favor of D38290 and D38291.

Pick nit in sparse_only_hole case.

Remove test cases that make no sense without the -S option.

As noted in D38191, our cp(1) transparently and unconditionally preserves holes when copying sparse files.

Moved test cases to D38290.

Independently of the matter of sparse files, should we remove the mmap() code entirely? Even in the absence of copy_file_range(), there is no real benefit outside of a narrow range of file sizes — much narrower, I suspect, than the [0, 8MB] range set out in the code.

In D38280#869740, @markj wrote:

It's not obvious to me why we bother with the ifdefs. Most other utilities in the base system don't conditionalize SIGINFO support, but I guess it doesn't hurt.

Fix unused var in !TARFS_DEBUG case.

Incorporate reviewer feedback.