Expose simulate_until to C API

[markaren]

No description provided.

[markaren]

How about this PR now?

[kyllingstad]

I'm so sorry, it turns out I have completely misled you here.

I thought cse::execution::simulate_until() returned a cse::step_result, but it doesn't. Instead, it returns a boolean that specifies whether the simulation actually progressed to the requested target time. A false return value does not indicate error, it only means that cse::execution::stop_simulation() was called.

So cse_execution_simulate_until() should probably also return a boolean. C doesn't have a boolean type¹, one just uses integers 0 and 1. So the function would return -1, 0 or 1 for error, false or true, respectively. Defining CSE_FALSE and CSE_TRUE constants in cse.h might be a good idea.

¹ At least not the version of C that is a subset of C++. Newer versions have the <stdbool.h> header, but we should probably stay clear of that.

[markaren]

How about the 2 last commits?

[kyllingstad]

Great!

The CSE_FALSE and CSE_TRUE constants will also come in handy when we add support for boolean variables to the C header. :)

[eidekrist]

So cse_execution_simulate_until() should probably also return a boolean. C doesn't have a boolean type1, one just uses integers 0 and 1. So the function would return -1, 0 or 1 for error, false or true, respectively. Defining CSE_FALSE and CSE_TRUE constants in cse.h might be a good idea.

1 At least not the version of C that is a subset of C++. Newer versions have the <stdbool.h> header, but we should probably stay clear of that.

Hold your horses! We have already included and used the <stdbool.h> header in cse.h. Should that be removed again? Otherwise, CSE_FALSE and CSE_TRUE are redundant.

The CSE_FALSE and CSE_TRUE constants will also come in handy when we add support for boolean variables to the C header. :)

Hold your horses! We already have support for retrieving/observing and overriding boolean variables :)

[markaren]

So I should just return the bool I get, and remove CSE_TRUE/FALSE ?

[kyllingstad]

@eidekrist:

Hold your horses!

:D My embarrassingly lacking knowledge of the cse.h header is becoming more and more apparent. Excuse me while I go find a dark corner in which to sit in shame.

We have already included and used the <stdbool.h> header in cse.h. Should that be removed again?

No, I guess it's fine. I considered that before replying to @markaren, but that header is not included in the "C subset" of C++¹, so anyone trying to use the cse.h header from C++ might run into problems. Then again, let's not be completely reactionary and only stick to ancient C standards.

New suggestion, then: Use the true and false constants from <stdbool.h>, but wrap the #include directive like so:

#ifndef __cplusplus
#    include <stdbool.h>
#endif

(I know, that part is strictly unrelated to this PR, but we might as well fix it now.)

¹ More precisely, the <stdbool.h> header from C99 and the corresponding <cstdbool> C++ header were introduced in C++11, deprecated in C++17, and removed in C++20. The reason is that they only define the bool type and the true and false constants, which are built-in keywords in C++.

[kyllingstad]

Note, the simulate_until function should still return int because it also needs to be able to return -1.

[markaren]

Ok now?

[kyllingstad]

Reaffirming approval. :)