RFC: Add tests for IPv6 fragmentation

[miri64]

With IPv6 fragmentation and reassembly merged (RIOT-OS/RIOT#11623 and RIOT-OS/RIOT#11596 respectively) during the 2019.10 release cycle, I think it makes sense to extend the release specifications for packets of sizes (MTU, 2*MTU]. Specifically I would extend 03, 04 and 07 with one task each to ping with packets with e.g. 2 KiB payload size.

All weird corner cases should already be covered by the tests/gnrc_ipv6_ext_frag application (which should be covered by 02, but who knows, since it requires sudo and people tend to skip tests that are not auto-run ;-P). However, I prefer to have at least some "in-the-wild" tests done during the release testing.

TBH, I personally never tested IPv6 + 6LoWPAN fragmentation (but AFAIK @benpicco did during the testing of RIOT-OS/RIOT#11623 [edit]after rereading his review comment I'm less sure[/edit]), so we have to see if the 04 extension works as expected and I don't believe multihop (07) was properly tested yet (tests/gnrc_ipv6_ext_frag tests forwarding of fragments though). I would thus suggest that the 04 and 07 extensions should be marked as experimental at least for the first RC they are tested in, in case some weird side effects arise.

[leandrolanzieri]

+1 for extending the tests and making those tasks experimental for now in 04 and 07, I think we should test this feature in multiple conditions.

On a separate matter, tasks 4.(5-8) still remain experimental? They seem to have been passing, @aabadie ?

[miri64]

Ok, then I prepare a PR with the specifications.

[miri64]

Good that we test this. Pinging with 2KiB already maxes out the packet buffer xD

[fjmolinas]

+1 for extending the tests and making those tasks experimental for now in 04 and 07, I think we should test this feature in multiple conditions.

I agree +1

[miri64]

On a separate matter, tasks 4.(5-8) still remain experimental? They seem to have been passing, @aabadie ?

Let's discuss this in a separate issue, please.

[kb2ma]

I just merged #137. With @leandrolanzieri and @fjmolinas agreement here, let's close this issue.