09:43 <achambe> So, anyone have any applications they are writing in zig? or is it still early days

10:02 <Aequus> Err, using -DCMAKE_PREFIX_PATH didn't work, I get a silly error saying file INSTALL cannot copy file "..." to "/usr/local/bin/zig" even though I don't want it to install to /usr/local, that's why I set it in the first place... When I manually edit cmake_install.cmake and change /usr/local to the prefix I want, it installs it there, but trying to compile anything with Zig, it tells me that it cannot find zig

10:02 <Aequus> lib directory and to reinstall or use --zig-install-prefix, but when I pass --zig-install-prefix=PREFIX, it says invalid argument. Any ideas as to how I can install and use Zig properly from PREFIX?

10:03 <Aequus> Also when I run "make install", it says "Install configuration: "Debug"", how do I change it to release, if possible?

10:03 <Aequus> I know how to workaround the issue, but I want to do it the proper way. :P

10:04 <Aequus> Funnily enough it attempts to install the libembedded_lld_*.a to the PREFIX, but not the zig executable.

10:08 <Aequus> Actually nevermind, my problem was that I used = after --zig-install-prefix. >.>

10:08 <Aequus> .ls

10:08 <Aequus> "zig --zig-install-prefix ~/.local/pkg/zig-0.2.0 build-exe --release-fast hello.zig" works.

10:08 <Aequus> Sorry. :P

10:11 <Aequus> And as to why it tried to install the libembedded_* stuff to CMAKE_PREFIX_PATH whereas the rest to CMAKE_INSTALL_PATH, I dunno.

10:14 <Aequus> CMAKE_INSTALL_PREFIX*

10:19 <Aequus> Huh, I rebuilt and reinstalled with -DCMAKE_BUILD_TYPE=Release, and when issuing the same command that worked with DEBUG, I'm getting:

10:19 <Aequus> : CommandLine Error: Option 'color' registered more than once!

10:19 <Aequus> LLVM ERROR: inconsistency in registered CommandLine options

10:23 <tiehuis> Aequus: when you run your `cmake ..`, is it showing clang with .so files and llvm with .a or vice versa?

10:24 <tiehuis> also, if you want something easy, you can try the master builds from CI: https://ziglang.org/download/

10:24 <Aequus> tiehuis: Oh, it uses gcc.

10:25 <Aequus> Well it also says found LLVM and CLANG, so I dunno.

10:25 <tiehuis> these are all statically linked so you shouldn't have any particular issues

10:25 <tiehuis> i was having that issue show up in arch when building from source, what is your os/distro?

10:25 <Aequus> arch

10:26 <tiehuis> you can install the package the aur and that should work

10:26 <tiehuis> i had to add a patch to the build step to get around this particular issue

10:27 <Aequus> Ah, all right. Thanks. :)

10:27 <tiehuis> as it seems libclang doesn't provide .a files for static linking so when building it finds the libclang.so and llvm.a files and there is an issue when dynamically linking libclang but statically linking llvm

10:34 <Aequus> tiehuis: Heh, the only reason I wanted Release is because I thought it'd disable dumping a core when the .zig file I'm trying to build doesn't exist. Any ideas?

10:34 <Aequus> This:

10:34 <Aequus> unable to open './hello.zig': file not found

10:34 <Aequus> Aborted (core dumped)

10:36 <tiehuis> aur does a release build, if you wanted to do manually however, you should just be able to modify your CMakeLists.txt with this patch:

10:36 <tiehuis> https://aur.archlinux.org/cgit/aur.git/tree/force_dynamic_llvm.patch?h=zig

10:36 <Aequus> Yes, I did that.

10:36 <tiehuis> you shouldn't get a core dump file with that by default

10:37 <Aequus> That's odd. :/

10:37 <tiehuis> maybe you have them enabled by default, have you tried anything here? https://wiki.archlinux.org/index.php/Core_dump

10:39 <Aequus> I probably have it enabled, but I thought it wouldn't dump a core with Release (and that patch), and I'm not sure if it's normal behavior from zig. I should probably disable automatic core dumps though.

10:39 <tiehuis> i don't get core dumps even with a debug build

10:40 <tiehuis> it should be an os dependent thing and not visible a zig thing i believe

10:41 <Aequus> Yeah but why would it abort when it could just gracefully exit having not found the source file? I thought it was a Debug/Release thing where with Debug it would abort, and with Release it would just exit or something.

10:44 <tiehuis> it shouldn't really, let me make an issue and i can change that tomorrow

10:45 <tiehuis> actually, i'll just do this now

10:57 <tiehuis> there are actually a few things to consider with how this interacts with tests so it's not a completely trivial change

11:50 <Aequus> LOL, I told my friend about Zig, I made the claim Zig's comptime is epic and generates optimized/better ASM, and he didn't believe so he gave me an example of gcd written in C, told me to write it in Zig, and also told me the optimal ASM would be "foo: movl $120, %eax ret". GCC and Clang transformed the tail recursion into a loop and then inlined it, but didn't precompute the result, etc... whereas what I did

11:50 <Aequus> was: copy his code (could have translate-c'd it) add comptime params, then do: "zig build-obj gcd.zig --release-fast --emit asm" and the result was exactly the optimal asm! I guess I convinced him... He was too lazy to bother with the talk that mentioned it so had to do it this way. :D

13:01 <achambe> Yeah, the comptime feature is really something I want to explore

14:33 <andrewrk> Aequus, unless you're trying to hack on the compiler itself, you can use the statically linked CI build of master branch

14:33 <andrewrk> https://ziglang.org/download/

16:57 <andrewrk> alexnask, I figured out #894

17:00 <alexnask> @andrewrk Cool, what was the issue?

17:01 <GitHub42> [zig] andrewrk pushed 1 new commit to master: https://git.io/vpvJH

17:01 <GitHub42> zig/master 4a2bfec Andrew Kelley: fix linux implementation of self exe path...

17:01 <andrewrk> this ^

17:01 <andrewrk> sorry for the trouble. pretty annoying bug

17:03 <alexnask> ah so basically it would read garbage after the executable path? Pretty weird that it seemed to happen randomly though :p

17:27 <GitHub164> [zig] andrewrk pushed 1 new commit to master: https://git.io/vpvUp

17:27 <GitHub164> zig/master b5459eb Andrew Kelley: add @sqrt built-in function...

18:34 <xkern> do you have a book recommendation for writing compiler ?

18:49 <andrewrk> xkern, I recommend trying it without reading any books

18:50 <andrewrk> and if you get stuck you can read other compiler source code for inspiration

18:50 <andrewrk> once you have something basic working without reading anything, you will know what you want to read about

19:22 <GitHub136> [zig] andrewrk pushed 1 new commit to master: https://git.io/vpvqK

19:22 <GitHub136> zig/master a8d7942 Andrew Kelley: exit with error code instead of panic for file not found

19:22 <GitHub69> [zig] andrewrk pushed 1 new commit to master: https://git.io/vpvq6

19:22 <GitHub69> zig/master 859b10d Andrew Kelley: std.math.ln and std.math.exp use float strict mode...

19:22 <andrewrk> Aequus, ^ panic issue fixed (at least for the case you were hitting)

19:24 <Aequus> andrewrk: Thanks. :)

19:25 <andrewrk> alright, back to playing with kernel threads :)

20:45 <xkern> andrewk: thanks . i understood from your message you start and if you get stuck look up tutorials.

21:48 <alexnask> Is there a way to switch the linker zig uses?

21:49 <alexnask> Or should I compile to an object file and have some script do the linking?

21:50 <andrewrk> alexnask, when linking, it's a non-negotiable assumption that zig will use its own linker (which is LLD)

21:50 <andrewrk> however, exporting .o files is supported, as you mentioned

21:50 <andrewrk> what's your use case though?

21:50 <alexnask> Ok, I assume on windows the object files are COFF?

21:50 <andrewrk> corret

21:50 <andrewrk> *correct

21:51 <alexnask> I want to use clinker (http://www.crinkler.net/)

21:51 <alexnask> I'd like to see what binary size I can get down to using just some basic std functions

21:51 <andrewrk> ah, the goal is to get the binary size down

21:51 <andrewrk> see also https://github.com/zig-lang/zig/issues/531

21:51 <alexnask> Yes, I'm looking at writing some simple 4KB demo

21:52 <andrewrk> you can also consider running the final executable through upx

21:53 <alexnask> I'm planning on trying out packers, yes, I just want to see if I can get the standard tools people use with C++ running with zig

21:54 <alexnask> I think it will be good enough even without LLVM -Os

21:54 <andrewrk> yes in this case I recommend taking advantage of build-obj

21:54 <alexnask> I will try to contribute to that issue for the compiler though, sounds simple enough

21:55 <andrewrk> alexnask, it should be pretty straightforward. start with modifying https://github.com/zig-lang/zig/blob/master/src/zig_llvm.cpp#L103 and then add the command line options to propagate the option

21:56 <andrewrk> we'll probably want to figure out what clang sets OptLevel to for -Os as well

21:56 <alexnask> Will do

22:18 <GitHub166> [zig] andrewrk pushed 1 new commit to master: https://git.io/vpvnY

22:18 <GitHub166> zig/master b936064 Andrew Kelley: add @atomicLoad builtin...

22:44 <andrewrk> alexnask, another possibly relevant issue: https://github.com/zig-lang/zig/issues/54