register_fprime_config: STATIC enforcement checks `SOURCE` instead of `SOURCES`

[thomas-bc-autowot]

F´ Version	devel (a41532a)
Affected Component	cmake/API.cmake — register_fprime_config

---

Problem Description

fprime_add_config_build_target is meant to force a configuration module to STATIC when it supplies sources or autocoder inputs, but it tests the wrong directive name:

# cmake/API.cmake
function(fprime_add_config_build_target)
    set(ARGN_PASS ${ARGN})
    # Ensure library is STATIC when supplying SOURCE or AUTOCODER_INPUTS
    if (SOURCE IN_LIST ARGN_PASS OR AUTOCODER_INPUTS IN_LIST ARGN_PASS)

The directive users pass is SOURCES, so the first test never matches. A config module with SOURCES but no AUTOCODER_INPUTS and no explicit STATIC is therefore not forced static and follows BUILD_SHARED_LIBS like any other library. In-tree config modules all have FPP inputs, so AUTOCODER_INPUTS masks this; present since #3642 (378a7a3).

How to Reproduce

1. Register a config module with SOURCES only (no AUTOCODER_INPUTS, no STATIC/INTERFACE).
2. Generate with -DBUILD_SHARED_LIBS=ON.
3. Observe the target is a shared library.

Expected Behavior

The module is built STATIC, as the docstring states. Fix: SOURCE -> SOURCES on that line, plus a cmake/test case for a SOURCES-only config module under BUILD_SHARED_LIBS=ON.

[Moferanoluwa]

Confirmed this reproduces, and I have a fix + regression test ready on a branch. Holding off on a PR until the CCB weighs in, per the contribution process.

Reproduction. Configuring cmake/test/data/TestConfigDeployment with -DBUILD_SHARED_LIBS=ON and building library_global_config — the config module in cmake/test/data/test-config-library that supplies SOURCES and no AUTOCODER_INPUTS:

[WITHOUT-FIX] artifacts found:
    liblibrary_global_config.so
[WITHOUT-FIX] RESULT: SHARED (.so) <-- followed BUILD_SHARED_LIBS, the bug

[WITH-FIX] artifacts found:
    liblibrary_global_config.a
[WITH-FIX] RESULT: STATIC (.a)  <-- forced static, correct

So the diagnosis in the issue is exactly right: SOURCE never matches the directive callers actually pass, and AUTOCODER_INPUTS is what masks it for every in-tree config module.

Fix. One word, SOURCE -> SOURCES, plus a cmake/test case as suggested: a new CONFIG_SHARED_BUILD fixture that configures TestConfigDeployment with BUILD_SHARED_LIBS=ON and asserts library_global_config is a static archive and not a shared object. Conveniently library_global_config (added recently for the GLOBAL_IMPLICIT_DEPENDENCY tests) is already a SOURCES-only config module, so no new fixture data was needed.

The full pytest -k config run passes (16 passed), and the before/after above confirms the new test actually fails without the one-word fix rather than passing vacuously.

Branch: https://github.com/Moferanoluwa/fprime/tree/fix/config-sources-static-enforcement — happy to open the PR whenever this is approved.

🤖 Generated with Claude Code

[sjsreehari]

Hi @thomas-bc-autowot , I'd like to take this on if it's available.

I looked into #5970 and the fprime_add_config_build_target() implementation in cmake/API.cmake. The STATIC enforcement currently checks for SOURCE, while the actual CMake directive is SOURCES.

Because of this typo, a register_fprime_config() call using SOURCES without AUTOCODER_INPUTS or an explicit STATIC does not get automatically promoted to a STATIC library and can instead follow BUILD_SHARED_LIBS.

My fix changes SOURCE to SOURCES in the condition and updates the adjacent comment accordingly. The change is limited to those two lines, with no other behavior affected.

Happy to adjust the approach or add a regression test if needed.