00:24 <shritesh> bketelsen: I went down the WebIDL rabbit hole this weekend. I think we can automatically generate bindings to *all* Web APIs.

00:26 <shritesh> I'm personally looking forward to being able to write WebGL in Zig.

05:42 <dbandstra> is there any way to construct a union value using metaprogramming?

05:45 <dbandstra> like this `ComplexType{ .Ok = 42 };` but where "Ok" comes from a string or something at comptime

05:45 <daurnimator> dbandstra: you want either https://github.com/ziglang/zig/issues/1214 or https://github.com/ziglang/zig/issues/383

05:46 <dbandstra> ah the infamous reify...

05:48 <dbandstra> well, it's a bit smelly that i'm using unions in the first place. i'll stick to switch statement boilerplate for now and then try to get rid of the unions later. thanks daurnimator

06:03 <hryx> tyler569: I figured out what was making my tests panic whenever there was a parse error. I'ma idjot https://github.com/hryx/zig/commit/2e8d5806c32b35a94bc21ce0a43f20cc7a77c242

06:37 <tyler569> haha always the best realizations

06:49 <hryx> sure are! Welp, the floodgates are open now. went from 0 passing tests to 10 so far /(°o°)\

06:49 <tyler569> \o/

11:25 <tgschultz> dbandstra, there's also: https://gist.github.com/tgschultz/e168cffd69003e4c788b6fd2ab1aea9e

16:04 <andrewrk> alright, crazy week is over, back to a normal schedule for me

16:04 <andrewrk> I'm going to start with these pull requests

16:11 <halo> :D

16:32 <shritesh_> This is in the front page of HN: http://way-cooler.org/blog/2019/04/29/rewriting-way-cooler-in-c.html I think Zig would be the perfect language for this use case.

16:33 <andrewrk> 11,000 lines of rust replaced with @cImport

16:40 <shritesh_> andrewrk: Would it make sense to generate Zig function signatures like @cImport does at compile time using the build system? I'm trying to create bindings for Web APIs from WebIDL. Has this use case been considered or are there any example of something like this?

16:41 <andrewrk> I think it's fair to say that some use cases call for code generation. And I think the best place for that to happen is at the build system level yes

16:52 <shritesh_> Alright. Chrome is currently experimenting with WASM host bindings where WASM can directly talk to the browser without going through JS. They did a writeup on WebGL and it was like 5x faster. I think this is a better route to prepare for. Zig can really shine when the feature lands in all the browsers.

16:53 <andrewrk> that seems like the obvious endgame for wasm in the browser

17:03 <emekankurumeh[m]> sounds like the equivalent of syscalls for wasm

17:04 <shritesh_> Here's the writeup if anyone is interested https://docs.google.com/document/d/1BOGEs1DAl0G366nTWE_qfWsISFhpFeU5nk4Jpu3wL74/edit?usp=sharing

17:19 <donpdonp> im trying to call pub extern fn pipe(__pipedes: [*c]c_int) c_int; but i dont know how to declare a variable that satisfies the parameter.

17:19 <donpdonp> can someone give an example? ive tried 5 different ways and failed

17:20 <andrewrk> donpdonp, it wants a pointer to an array of 2 c_ints

17:20 <andrewrk> right? isn't that what pipe wants?

17:20 <donpdonp> yes thats right

17:20 <donpdonp> i just dont know how to tell zig that with the compiler scolding me :)

17:21 <donpdonp> with/without

17:21 <andrewrk> var fds: [2]c_int = undefined; const rc = pipe(&fds);

17:21 <donpdonp> thx!

17:21 <andrewrk> np

17:23 <donpdonp> var fds: [2]c_int = undefined; c.pipe(fds); => error: expected type '[*c]c_int', found '[2]c_int'

17:24 <donpdonp> whoops missed an &

17:24 <donpdonp> but still get an error.

17:25 <donpdonp> c.pipe(&fds); => error: expected type '[*c]c_int', found '*[2]c_int'

17:25 <donpdonp> (this is the linux pipe() system call. it fills the int array with two file descriptors)

17:26 <andrewrk> donpdonp, that's a bug, I can fix that real quick

17:27 <andrewrk> [*c] is supposed to let you cast from anything: https://ziglang.org/documentation/master/#C-Pointers

17:27 <donpdonp> i see.

17:30 <donpdonp> im using 0.4.0+98fa065d by the way

17:45 <donpdonp> looks like i can use this in the interim: var fds = try allocator.alloc(c_int, 2);c.pipe(fds.ptr);

18:06 <tgschultz> you should be able to avoid avoid the allocation: var fds: [2]c_int = undefined; const fds_ptr = @ptrCast([*c]c_int, &fds); c.pipe(fds_ptr);

21:07 <andrewrk> donpdonp, ^ yes, the bug is that you need that @ptrCast

21:26 <donpdonp> got it. thx.

23:21 <fengb> andrewrk: should I adapt https://github.com/ziglang/zig/pull/2336 to use the ABI only? I'm also happy to close it and start afresh since I learned a bunch through the process

23:22 <andrewrk> fengb, up to you. I think I'll take responsibility of setting up the libc side of things though. Lots can go wrong and I want to pay close attention

23:22 <andrewrk> I mean as far as the .h files and building libc.a from source

23:25 <fengb> Sure

23:31 <fengb> I'll take a look at just the C ABI then