05:47 <emekankurumeh[m]> zig's color on github is really close to rust's.

05:50 <andrewrk> indeed

05:51 <emekankurumeh[m]> zig is rgb(236, 145, 92) and rust is rgb(222, 165, 132)

10:50 <tiehuis> emekankurumeh[m]: Some other similar language colors: https://i.imgur.com/sSbyVoC.png

10:51 <tiehuis> i dont think this matters as much now since the spacer was added between languages

12:50 <peekazig> Hi! I'm not actually using zig because of lack of time but I like to have a look at its progress every once in a while. I also look at nim periodically but I think it's main weakness is the anti-social attitude of its developers. I just wanted to say thank you and kudos to the zig community and specially to @andrewrk for being always nice to everyone.

13:03 <wilsonk> peekazig: yep, I agree with your sentiments about the quality of the Zig community, and Andrew also, of course :). (I haven't worked with Nim, though, so I can only comment on this community!)

16:33 <Akuli> a random thought about std/testing.zig: please don't add a method for everything, like python's stdlib unittest module does

16:34 <Akuli> unittest has: assertFalse, assertTrue, assertRaises, assertWarns, assertLogs, assertEqual, assertNotEqual, assertAlmostEqual etc

16:40 <schme245> so, I have a big where the function only returns the correct value if I add `warn("");` as the last statement of the inner loop. if I remove the warn, the value is garbage. only happens with --release-fast. how does one go about debugging a problem like this? :D

16:40 <schme245> *so, I have a problem

16:59 <schme245> huh, it works on the latest master!

17:06 <Akuli> given this: const A = union(enum) { Number: u8, None, }; how should i create an "instance" of A.Y? this works, but is it the best style? const a = A{ .None = void{} };

17:06 <Akuli> s/A.Y/A.None/

17:12 <Akuli> the docs have an example of instantiating what is A.Number in my example, but not how to use the None value

17:19 <schme245> on closer inspection, the latest master didn't solve my problem, it just made my reproducible test case pass. the problem in the original code persists :/

17:20 <andrewrk> schme245, debugging --release-fast is tricky. there are a lot of approaches

17:20 <andrewrk> for starters we can eyeball the assembly, and see if valgrind reports anything

17:21 <schme245> my debugging skills for native code are basically non-existant, unfortunately. everything I do with Zig is a learning experience

17:22 <schme245> what do you suggest for viewing the assembly?

17:22 <andrewrk> this could be quite fun then :)

17:22 <andrewrk> is your problematic code open source?

17:22 <schme245> I can make it open source if you give me 5 minutes to clean it up :)(

17:22 <andrewrk> sounds good

17:23 <andrewrk> Akuli, the main point of std.testing functions is to help show why tests failed. You can see how expectEqual would do that because it could, for example, show a string difference in a useful way. I don't see how having both assertFalse and assertTrue would accomplish that goal

17:27 <andrewrk> Akuli, you can implicitly cast an enum tag into a union if the payload type is void. const a: A = A.None;

17:27 <andrewrk> (or just leave it as the enum value, which will then implicitly cast when you use it)

17:32 <Akuli> about std.testing: going with 5 or 10 or 20 different asserter functions will likely result in what happened in python, most people not using the std solution

17:33 <Akuli> in python, there are alternatives that give the same amount or more debug information with simpler syntax, but even if there weren't any, i wouldn't want to type something like 'self.assertIsInstance(x, SomeType)' if i can do 'assert isinstance(x, SomeType)' (in python)

17:34 <Akuli> and the "alternatives" i mentioned run the code in some special way so that from 'assert isinstance(x, blabla)' they notice that it's a type assertion, and display a descriptive assert fail message accordingly

17:36 <andrewrk> it's fine if people want to use assert - they just have to understand that assert invokes undefined behavior in release-fast and release-small modes

17:37 <andrewrk> or in tests one can use expect, which panics even in release modes

17:37 <Akuli> i'm not saying that a "special assert" (whatever it was called in zig) for tests is bad, i'm saying that 10 or 20 "handy utility functions" that call it is bad

17:39 <Akuli> currently zig has 3 of those, which is fine

17:46 <schme245> andrewrk, where you able to reproduce the issue?

17:47 <andrewrk> schme245, give me a few minutes - I got a response to a question on llvm-dev mailing list and I'm replying

17:48 <schme245> no stress! :) I have to run in 10 minutes but I'll be back later this evening

17:48 <andrewrk> ok, take care

17:59 <Akuli> i have a script that pulls and compiles the latest zig, it's amazing how much new stuff is added every day! i don't think i have never seen a project that's developed this fast

17:59 <Akuli> for just a couple days: 244 files changed, 4210 insertions(+), 3966 deletions(-)

18:00 <andrewrk> that's mostly https://github.com/ziglang/zig/issues/1304