nhdp: deprecate module

[miri64]

Contribution description

Every time I change something about GNRC's network interface or sock I have to touch this module. However, I suspect that it is not to used anywhere (in our repo there is only a build test for it) and don't really expect it to still work.

The last substantial change to it that was not caused by API changes or CI-related code clean-ups was 5fbafbf which is almost 5 years old and it does not even follow our current programming practices (the header is not in the default include path), so I think it is safe to say that it is pretty much unmaintained.

For this reason, I propose to deprecate it and remove it after the 2020.04 release.

AFAIK, it the protocol this module implements is required for MANET-related protocols. Is anyone still working on those?

Testing procedure

After

make doc

the module should show up in doc/doxygen/html/deprecated.html.

Issues/PRs references

None

[jcarrano]

You may want to add __attribute__ ((deprecated)); to the init function to ensure that potential users get a warning. See https://gcc.gnu.org/onlinedocs/gcc/Common-Function-Attributes.html#Common-Function-Attributes

[miri64]

You may want to add __attribute__ ((deprecated)); to the init function to ensure that potential users get a warning. See https://gcc.gnu.org/onlinedocs/gcc/Common-Function-Attributes.html#Common-Function-Attributes

Did not know about that. Maybe we should use this for other deprecated features as well.

[kaspar030]

You may want to add __attribute__ ((deprecated)); to the init function to ensure that potential users get a warning. See https://gcc.gnu.org/onlinedocs/gcc/Common-Function-Attributes.html#Common-Function-Attributes

Did not know about that. Maybe we should use this for other deprecated features as well.

please no! did anyone look at the output of our unittests lately?

[kaspar030]

I'm talking about the compilation output: https://gist.github.com/kaspar030/a6a41b1bb2ad7147a3867b1a3f4aedd3

[jcarrano]

Yes, it has warnings, so? The issue there is that we don't have any way of filtering the output of the CI. We should not hinder the user experience because of our tool limitations.

[miri64]

@kaspar030 weren't you also advocating for compile warnings for API changes like this in the past?

[kaspar030]

@kaspar030 weren't you also advocating for compile warnings for API changes like this in the past?

Yes, one per issue. I've been compiling and debugging unittests lately, as is, this (attribute((deprecated))) is unusable.

[miri64]

I think the issue with ubjson spamming the compilation with deprecation notes stems from the fact that it is tied to a type which is used quite often in those tests. If I, as @jcarrano proposed just tie it to the init function it should only appear once.

[miri64]

I just try it and if you are not happy with the output you can still NACK. ;-)

[kaspar030]

I just try it and if you are not happy with the output you can still NACK. ;-)

ok, let's see.

[miri64]

also added a change I forgot to push earlier.

[miri64]

Oops, fixed typo

[miri64]

(Reordering my own PRs in the build queue. The other one is a bit more important as it fixes a bug)

[cladmi]

scan-build reported a "use after free" in nhdp so a +1 to get rid of it if not needed #11852

[jcarrano]

Ok. If anyone has concerns we still have time until 2020.04. Please squash.

[miri64]

Squashed.