Re: On OpenBSD and if_freenameindex(3).

[Simon Josefsson]

Eric Blake <address@hidden> writes:

> It turns out that MirBSD has the same bug regarding if_freenameindex
> not being a function.  So I've gone ahead and documented that; we can
> update the documentation when we implement the net_if (replacement
> header) and if_freenameindex (function implementation) modules.
>
> Also, if the test fails, it should print to stderr, not stdout.

Thanks!

> Meanwhile, we have generally been writing the tests so that the header
> test only checks things guaranteed by the header replacement with no
> other function modules in use, such as macros like IF_NAMESIZE and
> types like struct if_nameindex; while leaving signature checks up to
> the modules that actually replace functions.  But until we have any
> function modules, then the header test is the right place to test
> signatures. So, when we actually implement an if_freenameindex module,
> we'll probably rearrange part of this test into a new test file.  But
> I've run out of time for implementing net_if and if_freenameindex
> today, so here's what I'm pushing now:

Yes, I agree.  I believe it is useful to test for the functions
temporarily, because that will tell us whether replacement modules for
the functions are actually necessary.  It seems that the answer is
already yes for if_freenameindex though, but I fear the situation may
even be more complicated than that.

/Simon