02:33 <andrewrk> nice work mikdusan

02:33 <andrewrk> looks like thanks to your work we'll have aarch64-macos tarballs on the download page shortly

13:06 <LewisGaul> I'm trying to set the `log_level` in a test module, but when invoking `zig test foo.zig` `@import("root")` doesn't seem to correspond to the module `foo.zig` ?

13:09 <LewisGaul> Or maybe there's something else I'm missing about why log output isn't shown during 'zig test'

13:16 <fengb> For zig test, the root corresponds to test_runner.zig

13:28 <LewisGaul> ok makes sense, so is there no way to set the log level?

13:44 <LewisGaul> Also, is there any way to do this? `if (isError(result)) ...` where result is an error union e.g. `result: anyerror!u8`

13:45 <dutchie> `if (result) |_| {} |err| { ... }`

13:45 <dutchie> oops, i mean `else |err|`

13:46 <dutchie> or you can use `_ = result catch { ... }`

13:46 <LewisGaul> Yeah so I tried that but this is invalid syntax: `if (p.parse(input_nt)) |_| testing.expect(false) else |_| {};`

13:47 <LewisGaul> it's also a lot of syntax for things I don't actual care about (both payloads and the entire else block)

13:47 <LewisGaul> I want to do something *on success* (fail) but just continue on failure

13:48 <LewisGaul> so I kind of want the opposite of the catch block

13:48 <LewisGaul> I wanted to use `testing.expectError()` but it seems to only work with a specific error rather than being able to catch an error from a set of errors

13:48 <xackus> I think I will make a PR to add isError to std.meta

13:49 <LewisGaul> that would be cool - any advice/ideas on how I can do it for now?

13:51 <xackus> your if statement seems correct

13:51 <xackus> maybe you have an error in a different spot

13:55 <LewisGaul> I think the problem is that the `{}` after the else is interpreted as a block rather than as a 'void' value?

13:55 <Nypsie> In your if statement you're discarding the error, which is not allowed.

13:56 <LewisGaul> it works if I replace `{}` with `testing.expect(true)`

13:56 <Nypsie> Also the ';' after the block is invalid

13:57 <xackus> discarding is allowed

13:57 <xackus> this works: if (std.math.absInt(@as(i8, 3))) |_| {} else |_| {}

13:58 <LewisGaul> ahh it's the semicolon because it was intended to act as a block, got it :)

13:58 <xackus> but the semicolon is invalid indeed

13:59 <LewisGaul> I get confused by the fact blocks don't have semicolons but things like struct definitions do

13:59 <Nypsie> xackus, ah my bad. I guess the else case itself is required but you're allowed to discard the value.

16:58 <LewisGaul> Just wondering, is it possible to drop in an alternative implementation of `test_runner.zig` ?

16:59 <txdv> What does the code say?

17:01 <LewisGaul> well I couldn't see any comments about it in that module itself, or in the cpp file that mentions it

17:03 <txdv> the error handling is really handling all the cases https://github.com/ziglang/zig/blob/master/lib/std/special/test_runner.zig#L18-L20

17:17 <andrewrk> mikdusan, aarch64-macos tarballs are up on the download page

17:17 <g-w1> LewisGaul: what is the use case for dropping it in?

17:18 <mikdusan> andrewrk: the .toml order. I was thinking that might have something to do with it

17:18 <mikdusan> nice. thanks

17:20 <LewisGaul> g-w1 I'm kind of just wondering what's possible with zig tests:)  I have a set of tests that are filesystem-based, as in iterate over a directory to find subdirs that correspond to testcases. I'm having to do that in a single "test" block, wondering if I should just be using a custom test runner with its own 'main' or something instead, and not

17:20 <LewisGaul> sure how trivially it would integrate with the build system

17:21 <g-w1> well test cases are provided to the test runner, so it can't find new test cases

17:21 <LewisGaul> also thinking maybe I'd want to override `test_runner.log()` since that gets picked up as the `root.log()` implementation

17:23 <LewisGaul> I guess this is more in the direction of integration tests than unit tests (although I can still imagine cases this could apply to UT)

17:29 <LewisGaul> Also musing on the idea of having test orchestrators written in user-space in the future with features such as writing logs to file, continue if a test fails and report all results at the end, select and filter tests by module/test description, ... Wondering what that would look like/if there's anything that would need to be added for stuff like

17:29 <LewisGaul> this to be possible

17:30 <g-w1> do you want the custom test runner to find the tests or just run them? the latter is very doable

17:31 <LewisGaul> oh sorry I wasn't clear - the subdirs contain JSON files which correspond to testcases, as opposed to the subdirs containing test zig code

17:32 <LewisGaul> so testcases are effectively dynamically generated I guess

17:33 <LewisGaul> I have them all being run from a single "test" block because that's the only way it can work, but there's no way to tell zig it actually corresponds to multiple testcases that I'm aware of, so I get "All 1 tests passed" at the end. Not really a big deal, but just feels like a setup the standard zig test isn't built for?

17:43 <g-w1> setup you own runner like src/test.zig does that uses std.Progress

17:46 <LewisGaul> a-ha - "// TODO: find a way to treat cases as individual tests (shouldn't show "1 test passed" if there are 200 cases)" :)

17:46 <LewisGaul> yeah seems similar to what I've done

17:47 <LewisGaul> my code is at https://github.com/LewisGaul/zig-nestedtext/blob/main/tests/testsuite.zig for what it's worth

20:18 <bzoop> random question - is specifying any target (with -Dtarget) mutually exclusive with the env var ZIG_SYSTEM_LINKER_HACK from telling zig using the native lld linker (for macos-aarch64) -- even if the target ends up being the same (native-macos, for example)?

20:19 <g-w1> i *think* the system linker hack is only needed for targeting macos from macos, not 100% positive though. also, i dont think lld is native, ld is

20:21 <bzoop> i was seeing this behavior when fiddling around (trying to link system libcurl) - https://gist.github.com/bzoop/8c537ab3ed2b58a83bf5abb0f522314a

20:22 <g-w1> i would file a bug in this case

20:23 <bzoop> oh snap. i just remembered --verbose-link, and sure enough... if i specify a target "zig ld" gets used. otherwise system ld.

20:24 <g-w1> might be a bug in the logic in src/link/MachO.zig

20:24 <bzoop> -Dtarget=native still uses ld, but -Dtarget=native-macos used zig ld

20:25 <bzoop> k. tnx @g-w1! i'll file a bug on the off chance it wasn't expected behavior.

20:26 <g-w1> https://github.com/ziglang/zig/blob/master/src/link/MachO.zig#L681-L721 it happens here

20:27 <bzoop> and line 638 sets the bool for the if block

20:28 <bzoop> so perhaps self.base.options.is_native_os isn't quite a match somehow :thinking:

20:28 <g-w1> maybe macos is baseline, but you have some different version?

20:29 <lemur> What's the best-supported IDE/Editor for zig? I'm kind of spoilt on JetBrains, but the plugin seems to be quite outdated....

20:29 <bzoop> the native triple is pretty weird looking: "aarch64-macos.11.2.3...11.2.3-gnu"

20:29 <bzoop> but the "zig targets" .native.os is "macos"

20:29 <bzoop> *is listed as

20:30 <xackus> any editor that supports language server protocol, then you can use https://github.com/zigtools/zls

20:31 <xackus> sometimes it crashes, but it's great

20:32 <g-w1> most editors restart crashing lsp servers anyways :P

20:32 <lemur> xackus: ok, great! Thanks.

20:33 <xackus> vscode stops after 5 crashes

20:42 <bzoop> @g-w1 I filed an issue, in case you were interested in following: https://github.com/ziglang/zig/issues/8499

20:42 <bzoop> and thanks for your help!