[addder]

What was wrong

Since 7c73bbb (merged to master in #49), EngineCallExpression.compile pushes the arguments of engine routine calls first-argument-first. The NWScript VM, the original compiler and compiled game scripts push the last argument first, so the first argument ends up on top of the stack. Scripts from the built-in compiler pass swapped arguments or abort the VM with a type error, and omitted default arguments are pushed after the supplied ones. For example, SetGlobalNumber("A", 7) compiled to CONSTS "A"; CONSTI 7; ACTION 581, 2; it should be CONSTI 7; CONSTS "A"; ACTION 581, 2.

TestNCSCompiler did not notice, because it runs the output through PyKotor's own Interpreter, which got a compensating args_snap.reverse() in the same commit; the two errors cancelled.

The fix

• EngineCallExpression.compile: iterate the arguments from last to first (param_index still indexes function.params) and replace the comment that described the wrong order.
• Interpreter.do_action: the first parameter is on top of the stack, so pop in parameter order and drop args_snap.reverse().

User-defined function calls (compile_jsr) are not changed.

Tests

Four byte-level regression tests in TestNCSCompiler, next to the other engine call tests. They compare compile_nss output with hand-assembled NCS bytes and do not use the Interpreter:

• test_enginecall_argument_order_bytes_set_global_number: SetGlobalNumber("A", 7)
• test_enginecall_argument_order_bytes_three_mixed_types: a 3-argument call (object, int, float)
• test_enginecall_argument_order_bytes_omitted_defaults: omitted default arguments
• test_enginecall_argument_order_bytes_action_argument: an action argument (DelayCommand)

All four fail on the unpatched source and pass with this change.

I also compiled one call to every K1 and TSL routine whose parameters are int, float, string or object: all 185 (K1) and 221 (TSL) now push last-first.

Test results

Scoped command from CONTRIBUTING.md (Windows, Python 3.12, Libraries/PyKotor/tests):

	passed	failed	skipped
master (ca61ce8)	1604	5	114
this branch	1608	5	114

The only difference is the four new tests. The same five tests in tests/common/test_decode_fallbacks.py fail on both (test_8bit_encoding_only, test_bom, test_no_valid_encoding_found_strict_errors, test_strict_error_handling_decoding_failure, test_with_BOM_included); they are not related to this change. ruff check reports the same 7 findings on the three touched files before and after, and ruff format --check is clean for them.

bleeding-edge has the same two code sites (its classes.py and interpreter.py differ from master, so this commit does not merge cleanly there).

Fixes #351

[ProjectSynchro]

Closing this for now, Please see: #354