fix!(codegen): reset function outputs and default omitted arguments - #1958

3 findings in 4m 31s for $0.73 between f7d42e8 (master) and faf4f52 (anbt/PRG-4940):

• P2 book/user/language/functions.md:68: The claim that every output starts at its initial value or zero is too broad: codegen deliberately does not reset variable-length-array, REFERENCE TO, or alias outputs. Qualify the guarantee with these exceptions so readers are not led to expect their caller-owned storage to be cleared.
• P2 book/user/language/methods-and-properties.md:38: This still tells readers they must supply every method parameter and says an omitted parameter without a default has no defined value. Method calls now allow any parameter to be omitted and codegen supplies its default or zero, so update this paragraph to describe that behavior.
• P2 src/codegen/generators/pou_generator.rs:982: Runtime output defaults are evaluated without a function context: a method declaring VAR_OUTPUT p : POINTER TO Fb := ADR(THIS^); END_VAR now fails code generation. Static trace (not executed): maybe_get_constant_statement returns this address-unresolvable initializer, and generate_variable_initializer evaluates it with new_context_free, causing the THIS branch to return Cannot use 'this' without context. The lowered stack initializer can handle this expression, but the new prologue rejects it before that initializer runs.

Build Artifacts

🐧 Linux

From workflow run

🪟 Windows

From workflow run

2 findings in 5m 10s for $0.72 between 5adf03e (master) and 75d231e (anbt/PRG-4940):

• P2 book/user/language/functions.md:68: The aliasing warning says the shared input/in-out will read as zero, but an output with a nonzero initializer is reset to that initializer before the body runs. Describe the value as the output's initialized value (or zero when it has none), otherwise this example of the new behavior is inaccurate.
• P2 book/user/language/methods-and-properties.md:38: The new claim that a method call may omit any parameter contradicts validation: src/validation/statement.rs:2484-2507 still requires every VAR_IN_OUT parameter and REFERENCE TO input, reporting E030 when one is omitted. For example, fb.m() with m declaring VAR_IN_OUT x : DINT fails compilation instead of receiving the documented initialized temporary; the same incorrect rule also appears in the updated technical codegen page (static trace, untested).

3 findings in 3m 35s for $0.75 between 10ead7b (master) and 5e0e6de (anbt/PRG-4940):

• P2 book/technical/participants/06-init.md:144: This says outputs are zero-filled even when they have an initializer, but codegen initializes such outputs directly to the declared value (for example, the output-default snapshot stores 20 without first storing zero). Describe the entry behavior as applying the initializer or zero, with constructor work as needed, rather than requiring a zero-fill first.
• P2 book/user/language/methods-and-properties.md:35: The new entry-reset behavior also applies to outputs of methods when the caller supplies storage, but this page only describes the temporary used when an output argument is omitted. State that a supplied method output is reset to its initializer or zero on entry (subject to the reference, alias, and VLA exceptions), since otherwise callers may expect an unassigned output to retain its prior value.
• P2 src/codegen/generators/pou_generator.rs:944: Struct output defaults such as VAR_OUTPUT out : S := (x := 7); END_VAR reach this new initializer call with the by-reference pointer type hint, not S. Lowering decomposes the initializer into field assignments, so it does not correct the original aggregate's hint, and generate_literal_struct rejects it with Expected Struct-literal before those assignments run. This prevents functions and methods with struct output defaults from compiling (verified by static trace, not executed).

@codex review

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

5 findings in 4m 26s for $0.76 between 10ead7b (master) and 321b3c7 (anbt/PRG-4940):

• P2 book/user/language/functions.md:68: The overlap warning says the callee would read the aliased variable as zero, but entry initialization writes the output's declared initial value when present. For an output initialized to a nonzero value, the stated result is wrong; say it is reset to its default (or zero).
• P2 book/user/language/functions.md:84: The compiler now initializes an explicitly empty by-reference argument (=>/empty :=) to the parameter's default or zero in a temporary, but the user-facing call guidance never explains this behavior. Document the empty-argument case so callers know that it is defined rather than an uninitialized value.
• P2 book/user/language/methods-and-properties.md:38: The new claim that a method call may omit any parameter contradicts validate_call, which still requires every VAR_IN_OUT parameter and every REFERENCE TO input. For example, calling fb.m() when m declares VAR_IN_OUT x : DINT; END_VAR fails with E030 instead of supplying a zeroed temporary, so the documented calling pattern does not compile; the same incorrect claim appears in the technical codegen page.
• P2 book/user/language/methods-and-properties.md:38: This describes omitted method outputs as temporary but omits that supplied VAR_OUTPUT arguments are also reset to their initializer or zero on every call. The function page documents this only for functions, leaving the newly introduced method behavior uncovered in the user-facing method guidance.
• P2 src/codegen/generators/pou_generator.rs:940: A struct-literal output initializer such as VAR_OUTPUT out : S := (x := 7); END_VAR is passed to the entry initializer even though it is already decomposed into lowered member assignments. Its resolved Assignment node retains the declaration's auto-pointer type hint, so generate_literal_struct rejects it with Expected Struct-literal before those assignments can run, preventing the function or method from compiling (static trace, not executed).

4 findings in 3m 32s for $0.70 between 10ead7b (master) and b54c8bd (anbt/PRG-4940):

• P2 book/technical/internals/00-pous.md:276: This says every function output is initialized through the caller's address, but REFERENCE TO, AT alias, and variable-length-array outputs are not reset. Qualify the statement with those exceptions.
• P2 book/technical/participants/06-init.md:144: The statement that codegen zero-fills every function/method output is too broad: it skips reset for REFERENCE TO, AT alias, and variable-length-array outputs. Qualify this initialization description with those exceptions.
• P2 book/technical/pipeline/05-codegen.md:210: The output-reset description excludes only variable-length arrays, but codegen also skips resets for REFERENCE TO and AT alias outputs. Add those exceptions to keep this explanation consistent with the user-facing function rules and implementation.
• P2 book/user/language/methods-and-properties.md:38: This says every supplied method output is reset, but codegen deliberately skips resets for variable-length-array, REFERENCE TO, and AT alias outputs. Add the same exceptions described in functions.md so method callers are not told these outputs are reset.

4 findings in 4m 48s for $0.58 between 10ead7b (master) and 15b7f84 (anbt/PRG-4940):

• P2 book/technical/internals/00-pous.md:276: The statement that each function output starts at its initial value or zero omits the REFERENCE TO and AT alias outputs that codegen deliberately does not reset. Qualify the statement with those exceptions (and the VLA exception).
• P2 book/technical/participants/06-init.md:144: This says every function/method output is zero-filled on entry, but codegen deliberately skips reset for REFERENCE TO and AT alias outputs as well as VLA outputs. Add those exceptions so the claimed initialization order matches the compiler.
• P2 book/technical/pipeline/05-codegen.md:210: The output-reset description only exempts VLA outputs, but codegen also leaves REFERENCE TO and AT alias outputs untouched. Add these exceptions to avoid promising a reset the compiler does not perform.
• P2 book/user/language/methods-and-properties.md:38: This promises that every supplied method output resets like a function output, but REFERENCE TO and AT alias outputs are not reset by codegen (and VLA outputs are also preserved). State the same exceptions as the functions page.

2 findings in 3m 57s for $0.54 between 10ead7b (master) and f7aad81 (anbt/PRG-4940):

• P2 book/technical/pipeline/05-codegen.md:210: This says every output is reset, with only VLA outputs excluded, but codegen also skips resets for REFERENCE TO and AT-alias outputs. The same overstatement appears in technical/internals/00-pous.md and technical/participants/06-init.md; name these exceptions there as well so the mechanism description agrees with the behavior documented for users.
• P2 compiler/plc_lowering/src/initializer.rs:265: Skipping every scalar output constructor loses address-valued type defaults, for example TYPE P : POINTER TO DINT := ADR(g); END_TYPE with VAR_OUTPUT out : P;. Static trace (not executed): DataTypeGenerator::generate_initial_value returns no initializer for pointer types, the new output-entry initializer stores null, and this guard prevents P__ctor(out) from applying ADR(g), so an empty function or method returns null instead of the declared default.

4 findings in 3m 51s for $0.67 between 46fc67d (master) and 45b2ab6 (anbt/PRG-4940):

• P2 book/technical/participants/06-init.md:144: This says every output is zero-filled and then gets a constructor call, but scalar outputs are initialized directly to their value and deliberately get no constructor call. Distinguish scalar outputs from aggregates and pointer outputs, for which constructors are applicable.
• P2 book/technical/pipeline/05-codegen.md:321: The omission rule incorrectly limits required REFERENCE TO parameters to inputs; method REFERENCE TO outputs are required too. State that no REFERENCE TO parameter of a method may be omitted.
• P2 book/user/language/methods-and-properties.md:38: This says only REFERENCE TO inputs must be supplied, but validation requires every REFERENCE TO parameter of a method, including outputs. Clarify that all method REFERENCE TO parameters are required.
• P2 compiler/plc_lowering/src/initializer.rs:257: Scalar outputs lose address-valued type defaults: with TYPE A : LWORD := ADR(g); END_TYPE and VAR_OUTPUT out : A, an empty function or method returns zero instead of ADR(g). Static trace (not executed): this guard suppresses A__ctor(out), while the entry reset finds no resolved constant for the address-valued default and stores zero; no variable-level initializer restores it. Keep runtime type-default initialization for scalar outputs and cover this case with a lit test.

1 finding in 4m 38s for $0.70 between 46fc67d (master) and b19b656 (anbt/PRG-4940):

• P2 book/user/language/functions.md:68: The unconditional “Every call starts the output” claim also covers external functions, but their C implementations have no generated entry code to reset the caller's output. Scope this guarantee to compiler-generated function bodies so callers are not led to rely on stale-value clearing across the C interface.

1 finding in 3m 37s for $0.73 between 94669ef (master) and c3b0ed2 (anbt/PRG-4940):

• P2 book/user/language/functions.md:92: This says a VAR_IN_OUT argument cannot be empty, but calls such as foo(inout1 := ) and prog(inout1 := ) are accepted and bind a temporary; this change initializes that temporary from the parameter/type default or zero. Document empty VAR_IN_OUT arguments as temporary-backed instead of declaring them invalid.