Hey just wanted to run a client using serde_someip against the CommonAPI - HelloWorld Example and ran into the issue that it by default uses utf16le.

On the server side of the default example I see a : [CAPI][ERROR] MethodWithReplyStubDispatcher (someip): deserialization failed! [4660.22136.30000.1] on the client side I specified: serde_someip::options::StringEncoding::Utf16.

So changing this to utf8 (in the example service fdepl) and with following SomeIpOptions in the client it works great:

pub struct VSomeIpDefaultOptions {}
impl SomeIpOptions for VSomeIpDefaultOptions {
    const BYTE_ORDER: serde_someip::options::ByteOrder =
        serde_someip::options::ByteOrder::BigEndian;
    const STRING_WITH_BOM: bool = true;
    const STRING_ENCODING: serde_someip::options::StringEncoding =
        serde_someip::options::StringEncoding::Utf8;
    const STRING_WITH_TERMINATOR: bool = true;
    const DEFAULT_LENGTH_FIELD_SIZE: Option<serde_someip::length_fields::LengthFieldSize> =
        Some(serde_someip::length_fields::LengthFieldSize::FourBytes);
    const SERIALIZER_USE_LEGACY_WIRE_TYPE: bool = true;
    const SERIALIZER_LENGTH_FIELD_SIZE_SELECTION: serde_someip::options::LengthFieldSizeSelection =
        serde_someip::options::LengthFieldSizeSelection::AsConfigured;
    const DESERIALIZER_STRICT_BOOL: bool = false;
    const DESERIALIZER_ACTION_ON_TOO_MUCH_DATA: serde_someip::options::ActionOnTooMuchData =
        serde_someip::options::ActionOnTooMuchData::Discard;
}

So I looked into the serde_someip implementation and compared it to the CommonApi Someip one and the issues I currently see.

• StringEncoding does not allow specifying a separate endianness compared to the overall BYTE_ORDER. So not sure how much sense this requirements makes but in this case the example seems to use a different endianness for string encoding.

• UTF16 String BOM seems to depend on endianness, but in serialize_str BOM always writes the BOM for UTF-16 BE: Wikipedia BOM capicxx-someip-runtime _writeBom

First of all, thanks for making this crate available, it's really awesome work. I really like the idea to base this on serde and also the convenience and readability it brings.

I am currently trying out this crate in a client that talks to a service that uses vsomeip. I assume this might be a common use case. So I thought I'd share the SomeIpOptions implementations that worked for me, hoping that it will relieve the next one who tries this from digging in the common someip runtime code.

This assumes common someip runtime defaults (_dep == nullptr) and may need some tweaking depending on ones custom configuration.

pub struct VSomeIpDefaultOptions {}

impl SomeIpOptions for VSomeIpDefaultOptions {
    const BYTE_ORDER: serde_someip::options::ByteOrder = serde_someip::options::ByteOrder::BigEndian;

    // https://github.com/COVESA/capicxx-someip-runtime/blob/0ad2bdc1807fc0f078b9f9368a47ff2f3366ed13/src/CommonAPI/SomeIP/OutputStream.cpp#L368
    const STRING_WITH_BOM: bool = true;

    const STRING_ENCODING: serde_someip::options::StringEncoding = serde_someip::options::StringEncoding::Utf8;

    // https://github.com/COVESA/capicxx-someip-runtime/blob/0ad2bdc1807fc0f078b9f9368a47ff2f3366ed13/src/CommonAPI/SomeIP/OutputStream.cpp#L228
    const STRING_WITH_TERMINATOR: bool = true;

    // https://github.com/COVESA/capicxx-someip-runtime/blob/0ad2bdc1807fc0f078b9f9368a47ff2f3366ed13/src/CommonAPI/SomeIP/OutputStream.cpp#L246
    const DEFAULT_LENGTH_FIELD_SIZE: Option<serde_someip::length_fields::LengthFieldSize> = Some(serde_someip::length_fields::LengthFieldSize::FourBytes);
    
    // https://github.com/COVESA/capicxx-someip-runtime/blob/0ad2bdc1807fc0f078b9f9368a47ff2f3366ed13/src/CommonAPI/SomeIP/OutputStream.cpp#L246
    const OVERWRITE_LENGTH_FIELD_SIZE: Option<serde_someip::length_fields::LengthFieldSize> = Some(serde_someip::length_fields::LengthFieldSize::FourBytes);

    const SERIALIZER_USE_LEGACY_WIRE_TYPE: bool = false;

    const SERIALIZER_LENGTH_FIELD_SIZE_SELECTION: serde_someip::options::LengthFieldSizeSelection =
        serde_someip::options::LengthFieldSizeSelection::AsConfigured;

    const DESERIALIZER_STRICT_BOOL: bool = false;

    const DESERIALIZER_ACTION_ON_TOO_MUCH_DATA: serde_someip::options::ActionOnTooMuchData = serde_someip::options::ActionOnTooMuchData::Fail;
}

I had one interesting case when using ActionOnTooMuchData::Discard that I ran into CannotCodeString("String muse end with 0 terminator") because I had specified the max_size in Rust too short and the terminator was basically discarded. Even worse the difference between max_size and actual size was only the terminator in my case, so setting STRING_WITH_TERMINATOR to false magically made the deserialization work ;). One could say that is the definition of Discard , I'm just leaving it here for the next one who runs into this.

The following code fails to compile with: no associated item namedSOMEIP_TYPEfound for structVecin the current scope associated item not found in Vec<u8>

    use serde::{Serialize, Deserialize};
    use serde_someip::SomeIp;

    type Bytes = Vec<u8>;

    #[derive(Serialize, Deserialize, SomeIp)]
    struct Foo {
       #[someip(id = 1, max_elements = 4096)]
        data: Bytes,
    }

Looks like the type definition is not supported because the struct definition

    #[derive(Serialize, Deserialize, SomeIp)]
    struct Foo {
        #[someip(id = 1, max_elements = 4096)]
        data: Vec<u8>,
    }

works just fine.

Tested with serde_someip v0.1.1.