Add cli.t for testing invocation of stow executable

Unlike the other tests, this actually treats stow(1) as a black box
script, running it directly rather than require-ing it as a library.
This allows us to check things like the exit codes returned.

MANIFEST

@@ -34,6 +34,7 @@ NEWS
 README.md
 t/chkstow.t
 t/cleanup_invalid_links.t
+t/cli.t
 t/cli_options.t
 t/defer.t
 t/dotfiles.t

t/cli.t

@@ -0,0 +1,50 @@
+#!/usr/local/bin/perl
+
+#
+# Test processing of CLI options.
+#
+
+use strict;
+use warnings;
+
+use File::Basename;
+use Test::More tests => 1;
+
+use testutil;
+
+#init_test_dirs();
+
+# Since here we're doing black-box testing on the stow executable,
+# this looks like it should be robust:
+#
+#my $STOW = dirname(__FILE__) . '/../bin/stow';
+#
+# but unfortunately it breaks things like "make distcheck", which
+# builds the stow script into a separate path like
+#
+#    stow-2.3.0/_build/sub/bin
+#
+# before cd'ing to something like
+#
+#    stow-2.3.0/_build/sub
+#
+# and then running the tests via:
+#
+#    make  check-TESTS
+#    make[2]: Entering directory '/path/to/stow/src/stow-2.3.0/_build/sub'
+#    dir=../../t; \
+#    /usr/bin/perl -Ibin -Ilib -I../../t -MTest::Harness -e 'runtests(@ARGV)' "${dir#./}"/*.t
+#
+# So the simplest solution is to hardcode an assumption that we run
+# tests either from somewhere like this during distcheck:
+#
+#    stow-2.3.0/_build/sub
+#
+# or from the top of the source tree during development.  This can be done
+# via the following, which also follows the KISS principle:
+my $STOW = 'bin/stow';
+
+`$STOW --help`;
+is($?, 0, "--help should return 0 exit code");
+
+# vim:ft=perl