nanocoap: make coap_get_block2() actually fill struct

[kaspar030]

Contribution description

Previously, coap_get_block2() did not fill the passed structure, making it unusable.

This refactors coap_get_block1() into a shared coap_get_block() function and implements both coap_get_block1() and coap_get_block2() using that.

Testing procedure

• ckeck if the coap test applications still work as expected

Issues/PRs references

[kaspar030]

@kb2ma I hope this doesn't collide too much with your stack of CoAP PRs. Could we get this in first? It is a dependency of SUIT updates.

[kb2ma]

This commit in this PR is similar to 0ff2856 as described in #11057, which is the last of the sequence to completely implement Block. Functionally, I have no concerns here. Unfortunately the implementations are not identical, so merging this PR adds to the bitrot there.

[kaspar030]

This commit in this PR is similar to 0ff2856 as described in #11057, which is the last of the sequence to completely implement Block. Functionally, I have no concerns here. Unfortunately the implementations are not identical, so merging this PR adds to the bitrot there.

I'm fine with reverting this PR's commit in #11057 as first step. That should prevent the bitrot?

[kb2ma]

Sounds good to me.

[kb2ma]

This PR is a straightforward extension of the existing block1 getter (server request handler) to also get block2 (client response handler). Tested block2 by running block client test app. Confirmed no regression with block1 by running nanocoap server for /sha256 resource.

In coap_get_block(), the blknum and szx variables are not necessary. However, this PR is intended to be a minimal impact stopgap until we merge #11057 in 2019.10, which addresses this issue.

[kaspar030]

Thanks for the quick review!