00:03 <presiden> ah, solved

00:21 <andrewrk> presiden, did the above code work?

00:22 <diltsman> andrewrk, I added the export to the variable holding the vector table. That did require the type of the vector table to conform to the C ABI.

00:27 <andrewrk> presiden, ah I see the problem. I'm going to push a commit that makes your code work verbatim.

00:28 <andrewrk> diltsman, here's the code I tried: https://clbin.com/YL1DH

00:30 <diltsman> andrewrk, that is working now. Not certain what I was doing differently before that the u32 didn't work there.

00:32 <presiden> andrewrk: I'm doing something along: comptime var n = 0; comptime { n = foo(10) }; var arr: [n]bool = undefined;

00:32 <andrewrk> presiden, pushed: https://github.com/ziglang/zig/commit/ef5d7ce46382fb2735e83958fe1bbe416624906a

00:33 <andrewrk> look at the new test case at the bottom

00:34 <presiden> andrewrk: nice :)

00:35 <diltsman> Looks like the last unexpected thing is that while(true){} generates a pair of consecutive unconditional branches that both branch to the second branch instruction.

00:37 <jjido> Is it worth optimising empty brackets ?

00:49 <andrewrk> diltsman, are you referring to debug mode assembly?

00:51 <diltsman> Whatever the default output is when mode = b.standardReleaseOptions()

01:04 <andrewrk> diltsman, I would make a case that the unnecessary branch instruction is harmless - debug mode favors fast compilation, not optimal machine code

01:04 <andrewrk> the default mode is debug

01:05 <andrewrk> does the unnecessary branch go away with -Drelease-fast ?

01:05 <andrewrk> that said, at some point I will take a look and see if there's some simple fix to emit better debug mode code. I'm doing quite a bit of that in the copy-elision-2 branch

01:12 <diltsman> andrewrk, I didn't realize that it was debug mode. Given what I understand about how loops tend to be transformed (naively) to assembly, the double branch wasn't too much of a surprise. However, I expected a few other inefficienies in debug mode, that is what threw me off.

01:12 <diltsman> -Drelease-fast eliminates the double branch.

01:25 <andrewrk> diltsman, I suspect you may be interested in the work I've been doing in the copy-elision-2 branch. Here's an example: https://i.imgur.com/keMRJ3r.png

01:26 <andrewrk> it's showing the LLVM IR, not assembly, but you can see on the right there are fewer alloca and memcpy instructions

01:26 <andrewrk> I'm hoping to get this merged within the next couple weeks

01:30 <andrewrk> this will significantly improve debug builds

02:36 <diltsman> andrewrk, I saw that (maybe a different image) on Twitter. Definitely has promise. Copy elision was a major improvement for C++11.

02:47 <presiden> andrewrk: related to previous discussion. should something along this be allowed too: var arr = []bool{false} ** foo(10);

02:55 <andrewrk> ah, presiden, I just opened an issue for that. It should be easy to solve: https://github.com/ziglang/zig/issues/1707

04:35 <diltsman> Took some stupid manipulation of file locations, but I finally got the Zig program on the Arduino Due and paused the debugger. Appears that the debug locations line up as expected.

06:33 <hryx> Can the weekly Twitch stream be advertised on ziglang.org and the GitHub wiki?

06:33 <hryx> I'm surprised I only found out about it yesterday while reviewing some chat logs

08:52 <dec05eba_> I noticed that if you have a public struct in a file and then a non-public function in the struct scope you can call the non-public function from another file that uses @import. Is this expected behavior?

16:24 <andrewrk> hryx, good idea. https://github.com/ziglang/zig/wiki/Community#other-stuff