Release v2.2.0: rename to NewPingPlus, Teensy 4 support, yield() fix

[jshaw]

Renames the library from NewPingESP8266 to NewPingPlus and prepares it for submission to the Arduino Library Manager.

Verification

Check	Before	After
Host tests (make test)	52 pass	67 pass
Teensy 4.0 / 4.1	compile error	OK
ESP8266 build	2 deprecation warnings	0 warnings
UNO / ESP32 / Teensy 3.2	OK	OK, 0 warnings
Async example (ESP8266/ESP32)	OK	OK, 0 warnings
arduino-lint --library-manager submit	clean	clean

Rename + sync with NewPing v1.9.7

• NewPingESP8266.{h,cpp} → NewPingPlus.{h,cpp}; class and macros renamed
• TRIGGER_WIDTH 12 µs, automatic one-pin mode detection, constructor drives the trigger pin LOW, protected members, PING_MEDIAN_DELAY 30000 µs
• Examples converted .pde → .ino, with 3.3 V divider and ESP8266 boot-pin warnings

Bug fixes found in pre-release review

yield() was compiled away on every platform. The header carried a #ifndef yield / #define yield() fallback, but every supported core declares yield() as a function, never a macro — so the guard was always taken and the empty macro replaced every yield() call in the library. The WDT/WiFi cooperation documented in v2.1.0 was silently inert. Verified on esp8266 3.1.2, esp32 3.3.8 and avr 1.8.8 with a #pragma message probe both before and after.

Teensy 4.x failed to compile. The i.MX RT1062 has 32-bit port registers that cannot be assigned to the volatile uint8_t* members the bitwise fast path uses (cannot convert 'volatile uint32_t*' to 'volatile uint8_t*'). Teensy 4 now uses the digitalWrite path; Teensy 3.x keeps the bitwise path.

ping_async() was non-functional in one-pin mode. The interrupt was attached before the trigger pulse, so with trigger_pin == echo_pin the library's own pulse fired the ISR, reported a bogus ~12 µs echo, and detached the interrupt before the real echo arrived. Two-pin behaviour is unchanged.

Also fixed:

• ICACHE_RAM_ATTR → IRAM_ATTR (deprecated on ESP8266 core 3.x), with a fallback for cores older than 2.5.0
• set_temperature() shifted the configured range by 1 cm by recovering the max distance from _maxEchoTime instead of storing it
• max_cm_distance + 1 overflowed 16-bit unsigned int on AVR, leaving the echo timeout at 0 so every ping returned NO_ECHO
• Config macros are #ifndef-guarded so -D build flags apply. A sketch-level #define can never work across translation units — docs corrected (it also emitted a spurious "redefined" warning in user sketches)

Packaging

• library.properties, .gitignore; examples/.DS_Store untracked
• convert_mm() added to keywords.txt and the README API table; set_max_distance removed from keywords (it is protected)
• CI workflow: host tests, arduino-lint, and a compile matrix over AVR, ESP8266, ESP32 and Teensy 3.2/4.0 — the Teensy 4 entry is what would have caught the blocker above

Testing

New regression tests cover the yield, set_temperature and clamping fixes. Each was mutation-tested: every fix was reverted in turn to confirm the suite actually fails. The first version of the clamping test passed against the broken code — the host's 32-bit unsigned int makes both orderings identical — so it was re-pointed at the now-observable _max_cm_distance member. The mock also counts yield() calls, so the compiled-away bug cannot return unnoticed.

Known limitation

ROUNDING_ENABLED=true remains an untested configuration — the test suite hardcodes truncation expectations and 11 tests fail under it. ping_mm() was made to honour the flag for consistency (it was the only converter ignoring it), but properly supporting the option needs the test expectations parameterized. The default (false) is unchanged and fully covered.

Not done here

• No tag yet — the Library Manager requires a semver tag. Tag v2.2.0 after this merges so it points at the merge commit.
• Hardware testing is incomplete. Only the Arduino UNO has been verified on real hardware; ESP8266, ESP32 and Teensy are compile-verified only. The yield() and one-pin ping_async() fixes are unverified on silicon.

🤖 Generated with Claude Code