More on SIGINT, longjmp, and getaddrinfo

[Florian Weimer]

* Hal Murray:

> Are there any test case names or addresses that are guaranteed to timeout?

As Carlos explained, we have tests awaiting integration into the
source that inject custom recursive resolvers.  The _res_init_fp patch
I posted recently intends to make that approach less hackish.

These tests already exposed the lack of TCP timeouts.

Timeout testing is fairly time-consuming (in terms of wall clock), so
I'm not sure if we can run the entire tests for every build.

> That seems like something that would be handy for regression testing
> so there might be an RFC describing a reserved name.

It's not because builders generally lack network access, and for those
who don't, there is too much variance.  You'd be testing recursive
resolver behavior, too.  Some issues only surface with hostile or
non-compliant resolvers, too.