align FirmwareInfo with esp_app_desc_t field size.

[taunusflieger]

This PR addresses the differences (some fields are too small, one field is too large) in field sizes described in issue esp-rs/esp-idf-svc#204. It ensures, that the fields in FirmwareInfo are large enough to hold the fields defined in esp_app_desc_t. For details see the issue above.

[MabezDev]

embedded-svc is meant to be completely unaware of esp-idf, so I don't think we should be changing this struct to fit esp-idf's alignment needs unless there is another valid reason to do so. Instead, I think that esp-idf-svc should be handling the conversion from the esp-idf type, to the svc type.

If it turns out the embedded-svc struct is unfit for its purpose, we can change it. Thoughts @ivmarkov @zRedShift?

[ivmarkov]

embedded-svc is meant to be completely unaware of esp-idf, so I don't think we should be changing this struct to fit esp-idf's alignment needs unless there is another valid reason to do so. Instead, I think that esp-idf-svc should be handling the conversion from the esp-idf type, to the svc type.

If it turns out the embedded-svc struct is unfit for its purpose, we can change it. Thoughts @ivmarkov @zRedShift?

I agree. If there is some field in the embedded-svc struct which is too short, let's extend it, but 100% alignment is not what we are seeking out here.

[4rzael]

The current implementation where the version field is 24 bytes long, leads to a bug in esp-idf-svc, where EspOta::get_running_slot fails when the partition's version is too long.

This can happen for example when using git tags with fairly long names (because esp-idf's version is based on git describe.

I'm not sure if this should be solved here, by increasing field sizes, or in esp-idf-svc, by truncating the fields if needed ?

[DaneSlattery]

This just caught me, esp-docs say the version string accepts 32 bits (including null terminator), but embedded-svc uses heapless::String<24>.

The APIs currently have a single implementation for the ESP32[-XX] / ESP-IDF.

I think given that this is ONLY currently supported by the esp, it would make sense to at least align to their implementation. Alternatively, esp-idf-svc could just truncate the version field.

Related:
esp-rs/esp-idf-svc#378
esp-rs/esp-idf-svc#411