Fix tests reporting (upstream)

[jimklimov]

Another part of work, done in NUT fork of libmodbus as a subset of networkupstools#3, ripe for up-streaming to minimize context differences for eventual RTU USB contribution as such.

This PR focuses on usability of the libmodbus test suite to investigate problems seen by make check in diverse platform builds (including FreeBSD, OpenBSD, a couple of illumos/Solaris distros, MacOS+HomeBrew, Linux, Linux+MinGW and MSYS2 for Windows target).

This selection of commits does not include any major attempts to fix the issues seen on those workers, but rather a large mostly cosmetic change set to represent the test progress:

• in tests/unit-test-client.c, a TEST_TITLE() macro was introduced to report the beginning of a test case, so that subsequent error messages issued by the library methods called by unit-test-client can be attributed to this test case and aid in troubleshooting. The title string is remembered as last_test_title[] and reported in subsequent success or failure of ASSERT_TRUE() message(s). Program text was reshuffled to begin the test cases with the title, and adapt the wording where needed, but there should not be any functional changes in the test cases just yet.
• as a separate commit spanning tests/unit-test.h.in and tests/unit-test-client.c, a FLUSHOUT() macro was introduced to facilitate debugging in Windows (GitBASH) console which tends to fluctuate between flushing out every character (so messing up reports from test server and client writing to same console) to buffering and only flushing every 4KB or so.
  • Anyhow, piping the outputs of client and server to dedicated cat's seems to have helped the situation a lot more than peppering the code with flushing or even configuring the behavior with setvbuf, as explored in the fork's original PR.
• in tests/unit-test-server.c, added clearer logging about reasons of test set-up failure, if it happens.
• finally, in tests/unit-test-server.c and tests/unit-test-client.c (with ip_or_device) and in tests/unit-test.h.in (with UT_BITS_ADDRESS_VAL), a few C fixes landed to help with some compilers and their view on standards and static analysis.

FYI: PR #792 seems like a useful addition to the test suite as well, although that case did not pop out in my experiments.

NOTE: CLA form posted a few weeks ago, ignored so far.

[cla-bot]

We require contributors to sign our Contributor License Agreement. In order for us to review and merge your code, please fill https://forms.gle/5635zjphDo5JEJQSA to get added. Your document will be manually checked by the maintainer. Be patient...

[cla-bot]

We require contributors to sign our Contributor License Agreement. In order for us to review and merge your code, please fill https://forms.gle/5635zjphDo5JEJQSA to get added. Your document will be manually checked by the maintainer. Be patient...

[cla-bot]

We require contributors to sign our Contributor License Agreement. In order for us to review and merge your code, please fill https://forms.gle/5635zjphDo5JEJQSA to get added. Your document will be manually checked by the maintainer. Be patient...