This makes the test faster and more robust in very a busy WiFi environment
(our CI) and still exercises the scenario of fragmenting the message on
both mqtt and ssl levels (binary size to send > mqtt buffer size)
esp_modem_stop_ppp() stops both ppp netif and switches the modem back to
command mode. IF these two actions are not synchronised, we might
experience issues of
* active PPP session trying to send/receive uart-data
* command mode already active before modem switched to it
both resulting in crashes.
Fixed by introducing the transition mode and running these actions in sequence
* set esp-modem to transition mode
* enter command mode, wait for the reply or re-sync
* close the PPP netif
* wait until the netif closes
Other fixes include ignoring certain events if modem component not ready
or not in appropriate mode:
* ignoring all UART events comming from DTE with no DCE attached
* ignore pattern detection in PPP mode
To be in line with other interfaces, when deleting the driver and it's glue layer to esp-netif, we DO NOT delete the related esp-netif (since is was allocated and created outside of the glue code and it works the same way in esp-eth and esp-wifi driver/glue layers).
Closes https://github.com/espressif/esp-idf/issues/5486
esp-modem netif glue layer implements esp-netif attach callback to setup corresponding driver functions and start the modem. The error code of esp_modem_start_ppp() was ignored and ESP_OK returned in all cases. Fixed by passing esp_modem_start_ppp()'s error code to post_attach callback.
Closes https://github.com/espressif/esp-idf/issues/5430
Add logic to set `FIN` flag automatically for transmitted WS frames, but
if `fragmented` option set indicating an expert/manual mode, then the
`FIN` flag is set according to the `final` option.
CoAP component relies on some mbedTLS crypto configuration
options, e.g. DTLS and PSK. These configuration options if
selected, have footprint impact on generic TLS examples like
https_request or https_ota as well.
Footprint of https_request example with/without change is per below:
$ ./tools/idf_size.py new_https_request.map --diff old_https_request.map
<CURRENT> MAP file: new_https_request.map
<REFERENCE> MAP file: old_https_request.map
Difference is counted as <CURRENT> - <REFERENCE>, i.e. a positive number means that <CURRENT> is larger.
Total sizes of <CURRENT>: <REFERENCE> Difference
DRAM .data size: 14796 bytes 14796
DRAM .bss size: 23560 bytes 23680 -120
Used static DRAM: 38356 bytes ( 142380 available, 21.2% used) 38476 -120 ( +120 available, +0 total)
Used static IRAM: 89045 bytes ( 42027 available, 67.9% used) 89045 ( +0 available, +0 total)
Flash code: 554231 bytes 563823 -9592
Flash rodata: 179000 bytes 181224 -2224
Total image size:~ 860632 bytes (.bin may be padded larger) 872568 -11936
This commit moves relevant config options to CoAP specific examples
and also adds some run time warnings if they are kept disabled.
Closes https://github.com/espressif/esp-idf/issues/5262
The interface name can be used in socket API, i.e. setsockopt(). The API esp_netif_get_netif_impl_name() shoudl be used to populate standard
interface structure struct ifreq.
If mdns answer hasn't been received within timeout, Value error would be raised, but the mdns-server-thread would still run, blocking CI jobs. Fixed by moving the raise statement within try-finally block
