gcoap: error signaling for gcoap_req_send() insufficient

[haukepetersen]

Description

Currently, gcoap_req_send() only has a single return value for all possible error cases. This is not sufficient, as it strips information that is mandatory for certain use cases of the API.

Example: flooding out large amounts of data. When sending a large amount of user data over CoAP as fast as possible (e.g. flood or throughput tests), one would e.g. use code like:

    do {
        res = gcoap_req_send(_txbuf, (hdr_len + len), remote, NULL);
    } while (res == ERROR_CODE_FOR_BUFFER_FULL);

Hoever, as gcoap_req_res() returns 0 for any kind of error, this code does break when for example exceeding the link layers MTU size with our payload -> sock_udp_send() returning -EMSGSIZE.

Steps to reproduce the issue

Run the code above for large payload sizes (payload exceeding you link capacity, e.g. > 1500 on native). The loop is never left.

Expected results

gcoap should return differentiated errors so the user can react accordingly.

Actual results

see above.

Versions

any :-)

[haukepetersen]

Oh, forgot to state: I though about creating a PR to improve this situation, but for many internal errors I can't (quickly) judge, how they can be best grouped. In the end I hope we agree, that we do not want to expose any possible error as unique value, but best group errors to a limited number or return codes, right?

[kb2ma]

I agree that the return value is the best way to provide this information, and gcoap should take advantage of it. Yes, I agree errors should be grouped, and I would think about identifying groups based on when a user may differentiate their response to the error.

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. If you want me to ignore this issue, please mark it with the "State: don't stale" label. Thank you for your contributions.