VHDL2019 assert api - partial

[Blebowski]

Add partial implementation of: LCS2016-050

Still need to look at tests, some are failing.

Escaped identifiers with leading underscores are used in std.env to minimize "exposure" of the
implementation specific function and shared variable (instead of e.g. assert_cnt).

[nickg]

This isn't the right way to implement it: the counters should be in C (e.g. in src/rt/stdenv.c) and incremented when x_assert_fail is called. The VHDL functions can then call into native code to read/clear the counters. Similar to how the directory and time functions in that file work.

[Blebowski]

OK, I will give it some more work. So the counters will be incremented by x_assert_fail. To access the counters from std.env.body, I should then use the foreign function that then links agains the C implementation (similarly as in e.g. file_line)?

[Blebowski]

OK, I see, similar to: ea457d5

[Blebowski]

Now ready for review

[nickg]

Suggested change

	return impl(warning) + impl(error) + impl(failure);
	return GetVhdlAssertCount(warning) + GetVhdlAssertCount(error) + GetVhdlAssertCount(failure);

[Blebowski]

What is the benefit of these two changes besides reducing number of places in the code where
the run-time is called? Isn't it going to be slower (altough negligibly, but still) to execute when
the "wrapping" function is called first instead of directly the function that returns the counter?
Or will the optimizations take care of that?

[nickg]

The overhead of the extra call is minimal and should be inlined anyway. This way avoid duplicating the symbol name / interface of the foreign function.

[nickg]

It's better to make the test self-checking and assert on the result of GetVhdlAssertCount, IsVhdlAssertFailed, etc. E.g.

assert GetVhdlAssertCount(warning) = 1;
assert IsVhdlAssertFailed;

[Blebowski]

That is true, but still the test needs to have "gold" report since the asserts are expected to fail. Return
code will not distinguish correct/incorrect

[nickg]

I added the format string part.