Fix Rolling build: fmt::runtime, Qt5/Qt6 dual-support, rviz Display::update, char8_t

[nbbrooks]

Why

Four Rolling/Resolute incompatibilities, all found by an end-to-end source build in an ubuntu:resolute container.

#	Problem
1	fmt v10+ made format strings consteval. Runtime-composed format strings in core/src/cost_terms.cpp and core/src/stage.cpp no longer compile.
2	rviz switched to Qt6. find_package(Qt5 REQUIRED) collides with a Qt6 rviz.
3	rviz's Display::update signature changed from (float, float) to (std::chrono::nanoseconds, std::chrono::nanoseconds) at RCLCPP_VERSION_GTE(30, 0, 0). TaskDisplay's override no longer matches, so the class stops being a concrete Display — the plugin fails to load and won't build with -Werror.
4	C++20 char8_t. u8"…" literals need reinterpret_cast<const char*> to feed Qt tr().

How

• fmt: wrap runtime-composed format strings in fmt::runtime().
• Qt: select Qt6 when building against rviz >= 15.1.14, else Qt5, with qt5_wrap_ui/qt6_wrap_ui switched to match. Gated on rviz_common_VERSION rather than Qt6 availability — see below.
• Display::update: dual-signature guarded on RCLCPP_VERSION_GTE(30, 0, 0); the Rolling branch takes nanoseconds, converts, and calls the existing implementation. humble/jazzy/kilted keep the old override.
• char8_t: reinterpret_cast<const char*> at each affected tr() call.

Why the Qt gate is on rviz's version, not Qt6 availability

find_package(Qt6 QUIET) succeeds whenever Qt6 is merely installed, which says nothing about which Qt rviz was built against — and this package links rviz_common/rviz_default_plugins, so its Qt must match. Ubuntu 24.04 ships both, so on jazzy or kilted with qt6-base-dev present the probe doesn't quietly mis-select — it aborts the configure, because rviz_common has already defined the versionless Qt::Core target:

CMake Error at Qt6CoreVersionlessTargets.cmake:42:
  Some (but not all) targets in this export set were already defined.
  Targets Defined: Qt::Core
  Targets not yet defined: Qt::CorePrivate

QUIET doesn't suppress that — it's a FATAL_ERROR inside Qt6's own config, not a not-found condition, so the Qt5 fallback is never reached.

The threshold is 15.1.14, not 15.1: ros2/rviz#1635 merged 2025-12-08 but first shipped in the 15.1.14 tag on 2025-12-17 — 15.1.0 through 15.1.13 are still Qt5.

distro	rviz	Qt
humble	11.2	Qt5
jazzy	14.1	Qt5
kilted	15.0	Qt5
lyrical	15.2	Qt6
rolling	16.0	Qt6

package.xml also needed fixing

visualization/package.xml requested qtbase5-dev unconditionally, which resolves to Qt5 on every platform — so rosdep would install Qt5 while CMake selects Qt6, which is precisely how the collision above arises. Switched to the versionless qt-base-dev, which rosdep maps per-platform (jammy/noble → qtbase5-dev, * → qt6-base-dev) — the same move ros2/rviz#1635 made in rviz_common's own package.xml.

Both fixes mirror PickNikRobotics/rviz_visual_tools#300, merged and green on all five distros.

Verification

Verified by an end-to-end source build of moveit_task_constructor (core + capabilities + visualization + demo) in ubuntu:resolute against a Rolling workspace. Every guarded branch preserves existing behavior on humble/jazzy/kilted.

Note the two Resolute CI jobs will fail on ros-{lyrical,rolling}-moveit-core not existing — moveit_core is not yet published for Resolute, which blocks every downstream package there. Unrelated to this change; those jobs are non-blocking.

🤖 Generated with Claude Code