03:59 <jbk> i think i have a lead on the failures.. when the net/ tests run, they are binding to localhost (not sure if that was intended vs. a non-loopback IP on the box).. but then when it goes to lookup the IP, it's ultimately calling getaddrinfo with the AI_ADDRCONFIG flag which (at least on illumos) excludes localhost addresses

04:00 <jbk> well binding to ::1

04:21 <slfritchie> I'm a big fan of Illumos & SmartOS. I haven't had an excuse to use it in anger since I left Basho. I'm looking forward to the porting work.

04:21 <slfritchie> On a different topic, I've discovered that `h.assert_eq[F64](w, 1.0)` will not fail a `ponytest` test if `w` is NaN.

04:23 <slfritchie> But `h.assert_eq[Bool]((w == 1), true)` will fail the test if w is NaN. Wheeee.

04:23 <slfritchie> (and h is a `TestHelper` object)

04:24 <slfritchie> This must be another corner of floating point behavior that I'm dreading learning about.

04:25 <jbk> i've been fortunate to have been able to largely avoid the sharp corners of floating point

05:22 <jbk> hrm.. ok so this is rather hairy

05:25 <slfritchie> (For the record, NaN needs special treatment. https://wiki.sei.cmu.edu/confluence/display/java/NUM07-J.+Do+not+attempt+comparisons+with+NaN)

05:25 <jbk> so when it creates the server socket, it uses getaddrinfo(NULL, ..) to get the 'any' address which always seems to return :: and 0.0.0.0, so then get it gets the socket address, and then maps that back to the loopback address

05:25 <jbk> (in map_any_to_loopback)

05:28 <jbk> except when it tries to connect, it uses getaddrinfo("::1", 0, ... with AI_ADDRCONFIG which explicitly excludes localhost

05:28 <jbk> which fails since this box does not have any non-IPv6 addresses configured

15:32 <jbk> though fbsd also returns PF_INET6/:: and PF_INET/0.0.0.0 as answers to the same getaddrinfo call

15:33 <jbk> so i wonder how other platforms behave if there's no non-loopback IPv6 addresses configured

18:14 <Shorttail> Where can I find the data transfered in a POST request? Is it in Payload? In the URL?

18:17 <SeanTAllen> POST data is a payload

18:58 <Shorttail> Thanks. Also, is there a good reason map lookups return V? instead of V | None?

19:28 <SeanTAllen> yes

19:28 <SeanTAllen> including... performance

19:42 <Shorttail> Oh. So, raising an error doesn't do a stack unwind? Nice

20:31 <SeanTAllen> No it does

20:31 <SeanTAllen> but that would be excepted to be a less likely case

20:31 <SeanTAllen> and if you are worried about it, you can do contains.

20:32 <SeanTAllen> additionally, you can intentionally store None in a map. That's a reasonable thing to do.

20:32 <Shorttail> Wouldn't using contains mean a double lookup in case it is there?

20:33 <SeanTAllen> Yes

20:33 <SeanTAllen> So, do you expect the key to be there or not?

20:33 <SeanTAllen> doing V | None would mean a match every time to get V or None from it.

20:33 <SeanTAllen> Which would also be slow every time.

20:34 <Shorttail> I get it. I'm just used to raising anything being slow

20:38 <Shorttail> About POST again, I can see Content-Length is greater than 0, but the body is empty. Where in the Payload is it then?

20:45 <SeanTAllen> well, it is slow, it just way faster than an exception in something like Java or C++

20:45 <SeanTAllen> you shouldnt throw error often in a hotpath

20:46 <SeanTAllen> Shorttail: i dont know the HTTP stuff and there's an RFC that will be approved as soon as I fix a few issues with it. That RFC will remove all the HTTP stuff from the standard library.

22:41 <brainproxy> compiling on macos, i get a a warning about llvm 6, but I do have 3.9.1 installed w/ brew

22:41 <brainproxy> do I need to set an env var or something before running make

22:41 <brainproxy> ?

22:47 <brainproxy> nvm, read the Makefile, which I should have done before asking :p

22:47 <brainproxy> export LLVM_CONFIG=/usr/local/Cellar/llvm@3.9/3.9.1_1/bin/llvm-config

22:49 <brainproxy> ^ works great!