CTests ctest.h has code that is not platform-agnostic. Also ctest sources use fopen() calls. Needs to be reworked.

This item was recognized during review of PR #212 by Alex. The points noted are:

One issue that you foresaw is that ctest.h doesn't adhere to our platform rules. I think that's out of scope, but we should open an issue and fix that.

About this chunk of code (which was inline in the .c code but macro'ized as a result of a review comment):

267 #define vfprintf_usermsg(fh, message)                \
268     do {                                            \
269         va_list varargs;                            \
270         va_start(varargs, message);                 \
271         vfprintf((fh), (message), varargs);           \
272         va_end(varargs);                            \
273     } while (0)
274

... the remark was: There is platform_error_log for this and I believe a way to specify the handle as well.
Revisit whether the call to vfprintf() on L271 can be changed to use platform_error_log() as suggested.

I want to note that this file is not platform-agnostic as is. Specifically, it uses some stuff that doesn't fly on certain platforms on which Splinter is also ported:

#include <inttypes.h> /* intmax_t, uintmax_t, PRI* /
#include <stddef.h> / size_t */
That seems straightforward to change using our platform definitions and some sed. This is the only ctest code file, right? We should open an issue and audit this. We have a special poison.h header to check for compliance.

A similar issue recognized during review of PR #224 is the following usages of fopen(), that is now being seen commonly in all unit-test sources:

108 // clang-format off
109 CTEST_SETUP(splinter)
110 {
111    Platform_stdout_fh = fopen("/tmp/unit_test.stdout", "a+");
112    Platform_stderr_fh = fopen("/tmp/unit_test.stderr", "a+");
113

Such references to fopen() should be platform'ized, so we don't run the risk that such code will not link on platforms where this stdlib function may not be available.