mbedtls-openipc: CMAKE_POLICY_VERSION_MINIMUM=3.5

[shipa-2]

Summary

• Set -DCMAKE_POLICY_VERSION_MINIMUM=3.5 so mbedtls-openipc configures on newer host CMake.

Test plan

• make BOARD=gk7102_lite (or any board using mbedtls-openipc) configures mbedtls without CMake policy errors

Part of splitting #2256.

Made with Cursor

[qodo-free-for-open-source-projects]

PR Summary by Qodo

mbedtls-openipc: set CMAKE_POLICY_VERSION_MINIMUM=3.5 for newer host CMake

🐞 Bug fix ⚙️ Configuration changes 🕐 Less than 5 minutes

AI Description

• Add a CMake policy minimum version to avoid configure failures with newer host CMake.
• Keep mbedtls-openipc build options unchanged while improving forward compatibility.

Diagram

graph TD
  A["mbedtls-openipc.mk"] --> B["CMake configure opts"] --> C["Host CMake"]

Loading

High-Level Assessment

The following are alternative approaches to this PR:

1. Patch upstream CMakeLists.txt (cmake_minimum_required / policies)

• ➕ Fixes policy behavior at the source for all consumers, not just Buildroot packaging
• ➕ Keeps policy logic closer to the project’s build system
• ➖ Requires maintaining and rebasing a local patch against upstream mbedtls
• ➖ May have broader behavioral impact than a packaging-only opt-in flag

2. Pin/limit host CMake version in the build environment

• ➕ Avoids modifying package configuration flags
• ➕ Can be applied globally in CI containers
• ➖ Does not help developers with newer local CMake
• ➖ Increases maintenance burden and reduces forward compatibility

Recommendation: Current approach is a good packaging-level fix: it resolves the immediate configure breakage on newer host CMake with minimal blast radius. Consider an upstream/patch-based solution only if more CMake policy issues appear that can’t be cleanly handled via configure flags.

Files changed (1) +2 / -1

Bug fix (1) +2 / -1

mbedtls-openipc.mkPass CMAKE_POLICY_VERSION_MINIMUM=3.5 to mbedtls-openipc CMake configure +2/-1

Pass CMAKE_POLICY_VERSION_MINIMUM=3.5 to mbedtls-openipc CMake configure

• Extends the package’s CMake configure options to set a minimum policy version of 3.5. This prevents configuration failures when building with newer host CMake versions that reject older policy defaults.

general/package/mbedtls-openipc/mbedtls-openipc.mk

[qodo-free-for-open-source-projects]

Code Review by Qodo

🐞 Bugs (0) 📘 Rule violations (0) 📎 Requirement gaps (0)

Great, no issues found!

Qodo reviewed your code and found no material issues that require review

Tip of the day

💡 Did you know, you can group findings by type and pick your Finding display, from Minimal to Full

More tips ↗ | Customize Qodo ↗ | Qodo docs ↗

[widgetii]

Thanks for splitting this out of #2256 — the diagnosis is right (mbedtls 2.25.0 really does die on CMake 4.x), but I believe this change is redundant: master already fixes CMake 4.x policy compat globally, for every package at once.

Already fixed in Makefile:7

export CMAKE_POLICY_VERSION_MINIMUM := 3.5

Added in 656094d — "Fix build with modern host toolchains (GCC 15, CMake 4.x)" (#1978). That line is present on origin/master and on this PR's own branch.

Three things I checked rather than assumed, with host cmake 4.3.3:

1. The failure is real. Configuring mbedtls-openipc-2.25.0 (cmake_minimum_required(VERSION 2.8.12)) bare → Compatibility with CMake < 3.5 has been removed from CMake. Hard error, as you say.
2. The env var alone is sufficient. Same tree, CMAKE_POLICY_VERSION_MINIMUM=3.5 cmake -S … -B … with no -D → Build files have been written to. CMake honours it as an environment variable, not only as a -D cache entry.
3. The export actually reaches cmake. Buildroot's pkg-cmake.mk CONFIGURE_CMDS sets PATH and the PKG_CONFIG_* vars but never scrubs the environment (no env -i), so the top-level export propagates into every target and host cmake invocation.

So the -D here duplicates a value cmake already has. Harmless, but not load-bearing.

It would also be incomplete on its own

Even without the global export, patching mbedtls alone wouldn't get a board to build. Packages using cmake-package whose sources declare min < 3.5:

package	cmake_minimum_required	boards enabling
mbedtls-openipc 2.25.0	2.8.12	93 ← this PR
libwebsockets-openipc 4.2.2	2.8.12	18 ← would fail next
mini / aura-httpd / usrsctp	2.8.12 / 2.8 / 3.0	0 (dormant)

Clean: aws-webrtc 3.6.3, libsrtp 3.21, baresip/libre 3.14, fdk-aac 3.5.1, yaml-cli 3.5, ipctool/jsonfilter 3.10. lzo (3.0) and libubox (2.6) are sub-3.5 cmake packages but enabled on no board. libzlib/libyaml/libcurl/libogg/opus have low minimums but build via generic/autotools infra, so cmake never runs on them.

That gap is itself the argument for keeping this global rather than per-package.

Minor

• The test plan checkbox is unticked and no checks have run on the branch yet — gcc-compat.yml triggers on general/package/**/*.mk, so it should fire once approved.
• Unrelated to this PR, but worth its own issue: mbedtls 2.25.0 is from Dec 2020 and long past EOL, shipping TLS on 93 boards.

---

If you are hitting a configure failure on current master, could you paste the error and your cmake --version? That would point at something the global export isn't covering, which would be worth fixing properly.

[widgetii]

This is already covered globally, so this line is a no-op in practice.

Makefile:7 on master has:

export CMAKE_POLICY_VERSION_MINIMUM := 3.5

(from 656094d / #1978). CMake honours CMAKE_POLICY_VERSION_MINIMUM as an environment variable as well as a -D cache entry, and Buildroot's pkg-cmake.mk CONFIGURE_CMDS doesn't scrub the environment — so the export already reaches this package's cmake invocation.

Verified with host cmake 4.3.3 against the extracted mbedtls-openipc-2.25.0 tree:

$ cmake -S mbedtls-openipc-2.25.0 -B /tmp/a
CMake Error at CMakeLists.txt:23 (cmake_minimum_required):
  Compatibility with CMake < 3.5 has been removed from CMake.

$ CMAKE_POLICY_VERSION_MINIMUM=3.5 cmake -S mbedtls-openipc-2.25.0 -B /tmp/c
-- Configuring done (0.7s)
-- Generating done (0.2s)
-- Build files have been written to: /tmp/c

The env-var form (which master already sets) is enough on its own.

[widgetii]

This thread is closed as no response has been received