Reconnect: keep the wait across short sessions, name a timeout and a lost link, setServer()

The wait between attempts started at 1 s again when the broker accepted the
session, so a session that was closed right after it opened (two boards with
one client id) came back every 2 s without end. It now starts again only
after a session that lasted 60 s; across shorter ones it keeps doubling.
An attempt given up after 30 s prints "the broker did not answer within
30 s", and loss of Wi-Fi prints "Wi-Fi is down, waiting for it" once per loss.
setServer(host, port) before begin() names another broker; the host is not
copied. basicLight: static RAM 30,616 B (+16), flash 333,701 B (+264);
91 host tests (bridge logic 50).

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
This commit is contained in:
David 2026-09-28 20:24:51 +00:00
parent d3e7ee7f2a
commit 42af413a6e
6 changed files with 164 additions and 23 deletions

View file

@ -194,9 +194,9 @@ for (AlexaDevice* lamp : lamps) {
## How it talks to Alex2MQTT
Everything goes over MQTT (port 1883 of `alex2mqtt.stormysdream.club`); the library makes no HTTP requests.
Everything goes over MQTT (port 1883 of `alex2mqtt.stormysdream.club`); the library makes no HTTP requests. `alexClient.setServer("broker.example.org", 1883)` before `begin()` names another broker, for example your own instance of Alex2MQTT. The host is a name or an address as text and is not copied, like the username and the password: pass a literal or a buffer that stays.
- **Session.** `begin()` starts SNTP (`pool.ntp.org`, `time.nist.gov`) and returns; `loop()` opens the MQTT session once Wi-Fi is up and the clock is set, or after 5 s of Wi-Fi without an answer, and subscribes to `<root>/discover` and `<root>/+/alexaDirective`. `getState()` is `CONNECTED` when the broker has acknowledged both subscriptions; `[Alex2ESP] error: the broker refused the subscription to <root>/discover` means that the root topic is not the one of the account. A session that ends or cannot be opened prints `[Alex2ESP] error: disconnected: <reason>; next attempt in N s` and is opened again after 1 s, then 2 s, 4 s and so on up to once a minute, for as long as Wi-Fi is up; the wait starts at 1 s again once a session is open. `the broker refused the username or the password` is the reason to look for when a board never shows up in Alexa. A sketch that sets the clock itself (its own `configTime()` with a time zone, an RTC) calls `alexClient.setTimeSource(false)` before `begin()`. The board has to reach an NTP server: DNS for the two names and outbound UDP port 123. Without the time of day the session still opens, the reports carry a `timeOfSample` in 1970, and the library prints `[Alex2ESP] error: the clock is not set, ...` when it connects and then at most once a minute while reports are sent.
- **Session.** `begin()` starts SNTP (`pool.ntp.org`, `time.nist.gov`) and returns; `loop()` opens the MQTT session once Wi-Fi is up and the clock is set, or after 5 s of Wi-Fi without an answer, and subscribes to `<root>/discover` and `<root>/+/alexaDirective`. `getState()` is `CONNECTED` when the broker has acknowledged both subscriptions; `[Alex2ESP] error: the broker refused the subscription to <root>/discover` means that the root topic is not the one of the account. A session that ends or cannot be opened prints `[Alex2ESP] error: disconnected: <reason>; next attempt in N s` and is opened again after 1 s, then 2 s, 4 s and so on up to once a minute, for as long as Wi-Fi is up. The wait starts at 1 s again after a session that lasted a minute: a session that the broker closes right after it has accepted it, as it does to one of two boards with the same client id, is opened again after the longer wait. `the broker refused the username or the password` is the reason to look for when a board never shows up in Alexa, `the broker did not answer within 30 s` stands for an attempt that the library gave up. When the link goes down the library prints `[Alex2ESP] error: Wi-Fi is down, waiting for it`, once per loss; the MQTT client reports the lost connection when its keep-alive runs out, and the session is opened again when Wi-Fi is back. A sketch that sets the clock itself (its own `configTime()` with a time zone, an RTC) calls `alexClient.setTimeSource(false)` before `begin()`. The board has to reach an NTP server: DNS for the two names and outbound UDP port 123. Without the time of day the session still opens, the reports carry a `timeOfSample` in 1970, and the library prints `[Alex2ESP] error: the clock is not set, ...` when it connects and then at most once a minute while reports are sent.
- **Discovery.** On `<root>/discover` the library answers with one discovery object per device on `<root>/discover_r`. The backend accepts one endpoint object per message and collects everything that arrives within 1 s for Alexa's discovery answer (up to 5 s for its proactive AddOrUpdate push), so all devices are published back to back from the next `loop()`. Each object sits on the heap (about 1 KB) until the MQTT client has sent it; when the client cannot take another one (free heap under 4 KB), the library prints `[Alex2ESP] discovery deferred at <endpointId>` and sends the rest from `loop()` as the queue drains, for up to 5 s after the request. `[Alex2ESP] error: discovery gave up: N device(s) not announced` means those devices missed this answer - on the backend's proactive discovery that can remove them from Alexa until the next one.
- **Directives.** The directive arrives as JSON on `<root>/<endpointId>/alexaDirective`. A directive larger than one TCP segment arrives in fragments, which are put together in one heap block that exists only until `loop()` has parsed it. `loop()` then calls the handler of the device: the one of `onDirective()`, or else `ReportState` or `Event` of `registerEvent()` (and before it `DirectiveReceived`, if registered) with the directive: `directive["header"]`, `directive["endpoint"]`, `directive["payload"]`. A device without a handler answers with the ErrorResponse `INVALID_DIRECTIVE`, and so does a device whose handler sent nothing for a directive that names a capability the device does not have; both print an error. A handler that sends nothing for a capability of its device leaves the directive unanswered, which prints `[Alex2ESP] error: <endpointId>: the handler sent no answer to ...`. Every call of `loop()` handles one directive, in the order of arrival. Up to eight directives wait for it: Alexa sends a group command ("turn off the kitchen") as one directive per endpoint, and they arrive faster than a busy sketch calls `loop()`. A directive that arrives twice (a broker that mirrors its topics delivers every message twice) is handled once; the repeat is recognised while it arrives and takes no place among the waiting ones.
- **Reports.** `send()` publishes the report on `<root>/<endpointId>/alexaResponce` at once. The backend waits 7 s for it, so answer from the event handler. Every property carries the board's UTC time as `timeOfSample`.
@ -282,10 +282,10 @@ Behaviour changes:
- Discovery is answered from `loop()`, not inside the MQTT callback. A discovery object over 3071 bytes (`ALEX2ESP_MAX_MESSAGE`) is refused with an error; the other devices are still announced.
- `getState()` stays `INITIALIZED` until the first connect, and becomes `CONNECTED` when the broker has acknowledged both subscriptions (1.1.0: the first of them). A subscription that the broker refuses prints an error that names the topic (without its root) and what to check, the root topic passed to `begin()`; the state stays `SUBSCRIBING`. When the MQTT client does not take a subscription, the session is closed and the next one subscribes again.
- Serial output goes through one log with levels. `alexClient.setLogLevel()` takes `AlexaLogLevel::NONE`, `ERROR`, `INFO` (the default) or `DEBUG`; `-DALEX2ESP_LOG_MAX=<0..3>` in `build_flags` sets the highest level that is compiled in (default 2, `INFO`). This replaces the `Alex2ESP_DEBUG` define inside `AlexaUtils.cpp`. 1.1.0 printed the memory figures for every MQTT message and the whole directive, correlation token included, for every directive; both are gone. A sketch or a build flag that defines `DEBUG`, `ERROR` or `INFO` as a macro (`#define DEBUG 1`, `-DDEBUG`) still compiles: the library sets those macros aside where it declares its levels and passes a level by its number everywhere else.
- The session is opened again whatever ended it, and the reason is printed in words: `disconnected: the broker refused the username or the password, check the two passed to begin(); next attempt in 1 s`. 1.1.0 reconnected every 5 s after a lost TCP connection only and printed nothing; a board with a wrong password, or one that met the broker while it was restarting, stayed offline until it was reset. The wait is 1 s after a session ended and doubles with every attempt that fails, up to 60 s. Nothing is tried while Wi-Fi is down, and the first connect waits for Wi-Fi as well (`waiting for Wi-Fi: ...` is printed once). A connect that has no answer after 30 s counts as failed. The MQTT keep-alive is 30 s (1.1.0: the 15 s of the MQTT client).
- The session is opened again whatever ended it, and the reason is printed in words: `disconnected: the broker refused the username or the password, check the two passed to begin(); next attempt in 1 s`. 1.1.0 reconnected every 5 s after a lost TCP connection only and printed nothing; a board with a wrong password, or one that met the broker while it was restarting, stayed offline until it was reset. The wait is 1 s at first and doubles with every attempt, up to 60 s; it is 1 s again after a session that lasted a minute, so a session that the broker accepts and closes at once does not bring the board back every second. Nothing is tried while Wi-Fi is down, and the first connect waits for Wi-Fi as well (`waiting for Wi-Fi: ...` is printed once). Loss of Wi-Fi prints `Wi-Fi is down, waiting for it`, once per loss. A connect that has no answer after 30 s counts as failed and prints `disconnected: the broker did not answer within 30 s`. The MQTT keep-alive is 30 s (1.1.0: the 15 s of the MQTT client).
- `getDevice()` before `begin()` prints an error: the device would have no root topic. A second `begin()` is ignored with an error.
- A report is sent through the client that created its device. An `AlexaDevice` or an `AlexaStatusMessage` that a sketch constructs itself has no client: `send()` returns `false` and prints `report for <endpointId> not sent: its device was not created by getDevice()`, where 1.1.0 queued the report. Use `getDevice()` and `buildStatusMessage()`. Both constructors take the client as an optional last argument, an `AlexaTransport*`.
- New: `Alex2ESP::setLogLevel()`, `Alex2ESP::setTimeSource()`, `AlexaDevice::hasEndpointId()`, `AlexaLog`, `AlexaSendResult`, and `AlexaTransport`, the interface a device sends and stamps its reports through. `Alex2ESP` implements it with `publish(topic, document)`, which sends a `JsonDocument` on the session of the library under the same checks as a report and returns an `AlexaSendResult`, and `timestamp(buffer, size)`, which writes the current time as `timeOfSample` has it.
- New: `Alex2ESP::setServer(host, port)` before `begin()`, for a broker other than the default; the host is not copied. `Alex2ESP::setLogLevel()`, `Alex2ESP::setTimeSource()`, `AlexaDevice::hasEndpointId()`, `AlexaLog`, `AlexaSendResult`, and `AlexaTransport`, the interface a device sends and stamps its reports through. `Alex2ESP` implements it with `publish(topic, document)`, which sends a `JsonDocument` on the session of the library under the same checks as a report and returns an `AlexaSendResult`, and `timestamp(buffer, size)`, which writes the current time as `timeOfSample` has it.
- Removed: the queues and buffers of `AlexaUtils` (`enqueue`, `dequeue`, `dequeueVals`, `enqueueReceive`, `dequeueReceive`, `isQueueEmpty`, `isQueueFull`, `isReceiveQueueEmpty`, `isReceiveQueueFull`, `receivePayload`, `nextMessageId`) and its `log`/`logln`, which printed nothing unless the library was edited; `AlexaUtils::printMemoryInfo()` stays. `MAX_STATUS_REPORT_SIZE` (the limit is `ALEX2ESP_MAX_MESSAGE`). The library no longer includes `ESP8266HTTPClient`.
- Interfaces are described by rows in program memory, `AlexaInterfaces::PowerController` and 27 more (see Interface Types), in place of the `switch` tables of `AlexaInterfaceUtils`, whose strings took RAM in every sketch. `addCapability()` takes a row or, as before, an `AlexaInterfaceType`; a sketch links only the rows it names. The discovery objects of the five examples are the same byte for byte.
- Versions and properties follow the interface pages of Alexa: `EndpointHealth` is announced as 3.1 (1.1.0: 3.3) and `ThermostatController` as 3.2 with its four properties (1.1.0: 3, none). `Speaker`, `StepSpeaker`, `PlaybackStateReporter`, `InventoryLevelSensor` and `WakeOnLANController` are announced as 3 (1.1.0: 1), `SimpleEventSource` as 1.0. The interfaces that 1.1.0 announced with an empty list of properties have their properties (`lockState`, `detectionState`, `mode`, `rangeValue`, `percentage`, ...). `SceneController`, `DoorbellEventSource`, `StepSpeaker` and `SimpleEventSource` are announced without a `properties` object.
@ -302,9 +302,9 @@ Behaviour changes:
- `PlaybackController` and `WakeOnLANController` are announced with `"properties": {}`, as their pages show them.
- Removed: `AlexaInterface::getJSON()` (`toJson()` adds the capability to the capabilities of its endpoint) and `getProps()` (the row has the properties), `ActionMapping::getJSON()` and the `String` and `std::vector` members of `ActionMapping`, the class `FriendlyName`.
Memory: `examples/basicLight.cpp` for a D1 mini takes 30,600 bytes of static RAM (1.1.0: 52,768) and 333,437 bytes of flash (1.1.0: 350,885), as PlatformIO reports them (espressif8266 4.2.1, Arduino core 3.1.2). 18,260 bytes of the static RAM were the five 2 KB queue slots, three more 2 KB buffers and the two HTTP clients; 3,720 were the names, versions and properties of all interfaces and the names of the display categories, which are in flash now. SNTP and the time stamp are 1.8 KB of the flash figure. The instance, names, configuration and mappings of a capability are 2.0 KB of it, the directive handler and the ErrorResponse 0.9 KB.
Memory: `examples/basicLight.cpp` for a D1 mini takes 30,616 bytes of static RAM (1.1.0: 52,768) and 333,701 bytes of flash (1.1.0: 350,885), as PlatformIO reports them (espressif8266 4.2.1, Arduino core 3.1.2). 18,260 bytes of the static RAM were the five 2 KB queue slots, three more 2 KB buffers and the two HTTP clients; 3,720 were the names, versions and properties of all interfaces and the names of the display categories, which are in flash now. SNTP and the time stamp are 1.8 KB of the flash figure. The instance, names, configuration and mappings of a capability are 2.0 KB of it, the directive handler and the ErrorResponse 0.9 KB.
Tests: `pio test -e native` in the repository runs 88 host tests: 47 of the receive and publish logic (reassembly of fragments, the directives that wait for `loop()`, repeated directives, the size limits, a heap without room, the wait between reconnects, topics, time stamps), 30 of discovery (the discovery object of every example against what 1.1.0 announced, the 28 rows against the interface pages, a type without a row, the discovery objects of a range, a mode and a toggle controller, a scene and a doorbell, what a capability refuses) and 11 of dispatch (the device and the capability a directive reaches, one handler for two devices, the handlers of 1.x, a capability more than a device holds, the answers to a directive for a capability the device lacks and to one without a handler). No board is needed.
Tests: `pio test -e native` in the repository runs 91 host tests: 50 of the receive and publish logic (reassembly of fragments, the directives that wait for `loop()`, repeated directives, the size limits, a heap without room, the wait between reconnects, topics, time stamps), 30 of discovery (the discovery object of every example against what 1.1.0 announced, the 28 rows against the interface pages, a type without a row, the discovery objects of a range, a mode and a toggle controller, a scene and a doorbell, what a capability refuses) and 11 of dispatch (the device and the capability a directive reaches, one handler for two devices, the handlers of 1.x, a capability more than a device holds, the answers to a directive for a capability the device lacks and to one without a handler). No board is needed.
Packaging: `library.json`, `library.properties` and the `softwareVersion` and `firmwareVersion` that a device reports in discovery say 1.2.0; the library has the number in one place, `ALEX2ESP_VERSION` in `src/AlexaVersion.h`. The `platformio.ini` of the repository is for the host tests; a sketch does not need it.