Hi @dawkagaming — thanks for working on system-library support! Distro packaging is a great direction. The CI Configure step is failing on build (Linux), check-windows, and analyze (cpp) — all three are blocked at CMake configure time before compilation starts. Here's what I see in the diff:

1. target_link_libraries(AetherSDR …) is called before the AetherSDR target exists

In the new zlib block (around line 63):

if (USE_SYSTEM_ZLIB)
    cmake_pkg_config(IMPORT zlib REQUIRED NAME zlib)
    target_link_libraries(AetherSDR PRIVATE PkgConfig::zlib)
else()
    add_subdirectory(third_party/zlib EXCLUDE_FROM_ALL)
    target_link_libraries(AetherSDR PRIVATE zlibstatic)
endif()

The AetherSDR executable target isn't created until much later in the file (via qt_add_executable(AetherSDR …) around line 870). CMake errors out with something like:

Cannot specify link libraries for target "AetherSDR" which is not built by this project.

This fires on the default configuration (USE_SYSTEM_ZLIB=OFF), which is why all three platforms fail — not just Linux. The original code worked because add_subdirectory(third_party/zlib EXCLUDE_FROM_ALL) just declares zlibstatic as an available target, and the actual target_link_libraries(AetherSDR PRIVATE … zlibstatic …) happened in the big bundled block at line ~932, after AetherSDR is defined.

Fix: Keep add_subdirectory(...) / find_package(...) early, but defer the target_link_libraries(AetherSDR …) calls until after qt_add_executable(AetherSDR …). The same pattern applies to your new mspack, libmosquitto, and alsa blocks.

2. cmake_pkg_config is not a standard CMake command

You're using cmake_pkg_config(IMPORT zlib REQUIRED NAME zlib) in several places. I don't think this command exists in vanilla CMake — once you fix issue #1 and CMake gets past the target ordering error, this will fail next with:

Unknown CMake command "cmake_pkg_config".

The standard pkg-config pattern in CMake is (docs):

find_package(PkgConfig REQUIRED)
pkg_check_modules(zlib REQUIRED IMPORTED_TARGET zlib)
# then later:
target_link_libraries(AetherSDR PRIVATE PkgConfig::zlib)

There's already precedent for this in the file — see the ENABLE_BNR block around line 277 which does exactly this for grpc++.

The alsa branch on line ~1041 also uses cmake_pkg_config and is unconditional on Linux (no USE_SYSTEM_* guard), which would break the Linux build outright even if the rest were fixed.

3. Minor: inconsistent option names

• Declared as USE_SYSTEM_LIBMOSQUITTO (line 11) but checked as USE_SYSTEM_MOSQUITTO (line ~1293) — these need to match.
• USE_SYSTEM_RTMIDI is checked but never declared via option(...).
• option(USE_SYSTEM_ZLIB OFF) is missing the description string — the signature is option(<var> "<help_text>" [value]), so right now OFF is being parsed as the help text. Not fatal but worth tidying.

Suggested approach: I'd recommend a smaller first PR that adds just USE_SYSTEM_ZLIB end-to-end with the correct CMake idioms — that gives reviewers a clean template, and then mspack/mosquitto/rtmidi can follow the same pattern in a follow-up. Happy to help if you'd like a paired-down version to start from.

Thanks again for tackling this — distro packaging support is genuinely valuable and the option-block consolidation at the top of the file is a nice readability improvement.

73 — AetherClaude 🤖

_{🤖 aethersdr-agent · cost: $9.8532 · model: claude-opus-4-7}

Actually the docs (https://cmake.org/cmake/help/latest/module/FindPkgConfig.html) mention that, quote:

The cmake_pkg_config() command for a modern and more advanced way to work with pkg-config in CMake without requiring pkg-config executable to be installed.

Here is the documentation for it: https://cmake.org/cmake/help/latest/command/cmake_pkg_config.html

Claude here — really nice work on this, Dawid! Distro packaging is exactly the kind of thing AetherSDR needs more attention to, and the USE_SYSTEM_* flag pattern you picked is the right shape for it. Saw your follow-up commit moving from cmake_pkg_config → pkg_check_modules(... IMPORTED_TARGET) — that's a great instinct; that command pair has been the durable cross-version CMake idiom for over a decade and works on every distro that matters.

Pushed a follow-up commit on top of your branch (with your permission via "allow maintainers to edit" — thanks for leaving that on). Three things in the merge:

1. Resolved the conflicts with current main. Two collisions, both clean to merge:

• Your centralised option block at the top of the file met the new LOWER_CASE_BINARY_NAME option (your other PR, Add option to make the binary's name lower case #3138 — already merged!) — combined them into the same block so all options stay in one place.
• Your new USE_SYSTEM_* target_link_libraries block met that PR's LOWER_CASE_BINARY_NAME OUTPUT_NAME logic just below — independent additions, kept both.

2. Dropped cmake_minimum_required from 3.25 back to 3.20. After your switch to pkg_check_modules(IMPORTED_TARGET ...) (which has been around since CMake 3.6), nothing in the new code actually needs 3.25 anymore. Restoring the original 3.20 keeps Debian 12 (which ships CMake 3.25) and older LTS distros buildable — the exact packagers this PR is meant to help.

3. Fixed an #ifdef HAVE_MQTT placement bug. This one was subtle: the previous form nested target_compile_definitions(... HAVE_MQTT) inside the if (NOT USE_SYSTEM_LIBMOSQUITTO) branch. So a packager building with -DUSE_SYSTEM_LIBMOSQUITTO=ON would link against the system mosquitto package — but HAVE_MQTT would never get defined, and every #ifdef HAVE_MQTT block in the rest of the project (MqttApplet, the radio model's MQTT plumbing, etc.) would compile out. Net result: a quietly broken MQTT build despite all the linkage being correct. Moved HAVE_MQTT to the outer ENABLE_MQTT scope so it gates the codebase regardless of bundled/system choice. Same fix for HAVE_MQTT_TLS — assumed on for system mosquitto unless -DMQTT_TLS=OFF (distro packages essentially always ship with TLS support).

Locally verified the default-bundled path (USE_SYSTEM_* all OFF) still builds clean. CI is re-running now.

This is a really useful contribution and I'm glad to land it — would love to see follow-ups for any other vendored dependencies you spot, and packaging recipes (AUR PKGBUILD, .deb control, etc.) if you end up writing any for your distro of choice.

Welcome to the project, and 73 from across the pond!

73, Jeremy KK7GWY & Claude (AI dev partner)

Claude here.

Merged at 10bc544. AetherSDR now supports opt-in USE_SYSTEM_ZLIB, USE_SYSTEM_MSPACK, USE_SYSTEM_LIBMOSQUITTO, and USE_SYSTEM_RTMIDI — exactly what distro packagers need for dynamic-linking-policy compliance. As a bonus, ALSA on Linux now goes through proper pkg-config discovery instead of the raw target_link_libraries(asound).

This is your fourth merged packaging-readiness PR (#3074, #3138, #3143, #3135), and reading them together they tell a coherent story: AetherSDR is meaningfully more packageable on Debian, Arch, NixOS, etc. than it was a week ago. Thank you for the persistent attention to that surface — distro-friendly software is rare and valuable, and you've moved a real measurable distance in that direction with minimal disturbance to existing bundled-build users.

Small follow-up on the same theme: I noticed that you closed #3139 (BUILD_TIMESTAMPS) in favour of the standard SOURCE_DATE_EPOCH approach we discussed. I just opened #3165 with that — a one-line CMake addition surfacing SOURCE_DATE_EPOCH honor at configure time, no source-tree changes. Credit to you in the PR body for surfacing the reproducibility concern in the first place.

73, Jeremy KK7GWY & Claude (AI dev partner)