Previously we used std::string_view, which pointed to the lower-layers
buffer which might have been reused for other asynchronous operations
before processing it, thus causing corruption of replies.
Closes https://github.com/espressif/esp-protocols/issues/463
ATD command is used to switch generic device to the data mode
by dialing GPRS_SRC. The generic version of the command was incorrect
`ATD*99##` (containing extra trailing # char, which is accepted by most
devices). Command has been fixed, but the extra #'s been kept for devices
that we cannot test (to be fixed in IDFGH-9470).
Closes https://github.com/espressif/esp-protocols/issues/433
at_raw() sends raw string and supports custom pass/fail phrases for the
API to return OK/FAIL. It also returns raw, unprocessed string as
output.
This allows sending custom requests, such as:
* dce->at_raw("", resp, "RDY", "...", 20000) -- waiting for reset
* esp_modem_at_raw(dce, "+++", resp, "DISCONNECTED", "ERROR", 5000) --
exits PPP mode
Closes https://github.com/espressif/esp-protocols/issues/471
Partially addresses https://github.com/espressif/esp-protocols/issues/468
If we received response in two chunks and the first one completes the
command (result=OK or FAIL), and the second chunk pre-empts command processing,
then we performed processing again rewritting the result back to
TIMEOUT. This would invalidate the command, but also causes an
exception: ESP_MODEM_THROW_IF_ERROR(ESP_ERR_INVALID_STATE)
Fixed by checking if the processing already finished in process_line().
Closes https://github.com/espressif/esp-protocols/issues/426
In the recent DTE refactoring (cb6e03ac) we install terminal callbacks
in the constructor, but the change missed initializing modem_state
in DTE constructors which take two terminals as arguments to work in
DUAL mode.
This change is needed for the fs terminal to work correctly again, after
merging 44bae24c78 which shifted
responsibility of thread safety from a terminal to the DTE. (Before it
was the terminal, which removed this callback safely when processing completed,
and thus we had to check the callback return value). Not the DTE layers
are responsible for thread safety and the return value indicates
success/failure from processing.
Also uses the default clock, the same way as UART terminal.
The callback should return false only if there is a problem with the received data.
The updated UART Terminal implementation cannot longer clear is callback.
It is DTE's responsibility to clear the callback.
* bool flag to defragment CMUX payload (useful for devices that send longer messages)
* int flag to force a delay between creating virtual terminals (useful for chatty devices that send some requests)
Race condtion:
* First command timeouted, but the reply came just after evaluation and
set signal variable to GOT_LINE
* Second command should timeout too, but a consistency check validates
that it timeouted and at the same time GOT_LINE (from previous step) and
throws an exception
STR:
* Revert change in esp_modem_dte.cpp
* Run TEST_CASE("DTE command races", "[esp_modem]")
Closes https://github.com/espressif/esp-protocols/issues/110
* Fix DISC message's control field (0xFF->0xFE) to keep Poll bit cleared
* Accept UIH resp on DISC for both (0xFF, 0xFE) ignoring P/F bit
* Fix timeout processing when handling DISC message
Closes https://github.com/espressif/esp-protocols/issues/103
