Oskar Hahn (Dec 01 2024 at 17:56): Oskar Hahn (Dec 01 2024 at 17:56):I changed the signature of my AoC-platform to Str -> Result Str _.
There seems to be a bug, when the compiler can not find out the type of the error value.
This works:
part1 : Str -> Result Str []
part1 = \input ->
input
|> Ok
This also works:
part1 = \input ->
input
|> Ok
|> Result.mapErr \_ -> ThisLineIsNecessaryForRoc
But this fails:
part1 = \input ->
input
|> Ok
thread '<unnamed>' panicked at crates/compiler/mono/src/ir.rs:6171:56:
called `Option::unwrap()` on a `None` value
stack backtrace:
Is this a known issue?
Eli Dowling (Dec 01 2024 at 18:25):if you run a debug build of the compiler, it will have a slightly more interesting error:
let mut iter_lambda_set = solved.iter_all();
debug_assert_eq!(
iter_lambda_set.len(),
1,
"{:?}",
(env.subs.dbg(*lambda_set), env.subs.dbg(ambient_function))
);
let spec_symbol_index = iter_lambda_set.next().unwrap().0;
That's the code causing the panic.
In debug it'll hit the assert and might give us some more info
Oskar Hahn (Dec 01 2024 at 18:38):With a debug build, I get the following error:
thread '<unnamed>' panicked at crates/compiler/mono/src/layout.rs:2066:17:
unspecialized lambda sets left over during resolution: LambdaSet([] + (<2418>FlexAble(*, [`Inspect.Inspect`]):`Inspect.toInspector`:2), ^<2421>), UlsOfVar(VecMap { keys: [2418], values: [VecSet { elements: [2416, 2420] }] })
Maybe it is relevant, that the platform runs this code:
part1ForHost : Str -> Result Str Str
part1ForHost = \input ->
part1 input
|> errToStr
errToStr : Result Str _ -> Result Str Str
errToStr = \r ->
r |> Result.mapErr Inspect.toStr
Yeah, the issue is Inspect.toStr probably
Brendan Hansknecht (Dec 01 2024 at 18:39):Cause it doesn't know how to inspect a *
Brendan Hansknecht (Dec 01 2024 at 18:40):Though I could be wrong, might be related to mapping a * error, but I recall that working.
Oskar Hahn (Dec 01 2024 at 18:40):Yes, you are right. You get the same error in the repl:
Ok "foo" |> Result.mapErr Inspect.toStr
By any chance does it work if you wrap Inspect.toStr with a lambda?
Oskar Hahn (Dec 01 2024 at 18:42):Brendan Hansknecht said:
By any chance does it work if you wrap
Inspect.toStrwith a lambda?
No, same error:
Ok "foo" |> Result.mapErr \err -> Inspect.toStr err
And I assume it works if you remove the Inspect.toStr? Just replace with identify function?
Oskar Hahn (Dec 01 2024 at 18:44):Brendan Hansknecht said:
And I assume it works if you remove the
Inspect.toStr? Just replace with identify function?
Yes. This works:
Ok "foo" |> Result.mapErr \err -> err
Yeah. So it just that you can't inspect a *. Hmm... I wonder if we can still autoderrive for that and add in a dummy value. At a minimum this should produce a good error about applying abilities to *
Oskar Hahn (Dec 01 2024 at 18:46):But why does it work in basic-cli? https://github.com/roc-lang/basic-cli/blob/6b9bc85d18c30cb34bf1351067fe3ec94811bf93/platform/main.roc#L44
Brendan Hansknecht (Dec 01 2024 at 18:46):Or maybe we can just make abilities not generate when called on *. Like it would be fine to generate a crash in those cases. Cause those cases are impossible to call.
Brendan Hansknecht (Dec 01 2024 at 18:47):Basic cli always has at least the Exit error tag
Brendan Hansknecht (Dec 01 2024 at 18:48):Oh, I bet you could work around this with a dummy tag
Oskar Hahn (Dec 01 2024 at 18:49):This does not seem to work. I get the same error, if I change the signature to part1 : Str -> Result Str [SomeError]_
Brendan Hansknecht (Dec 01 2024 at 18:49):Result.map \err ->
when err is
Dummy -> Inspect.toStr Dummy
e -> Inspect.toStr e
Yes. This works! Thank you
errToStr : Result Str _ -> Result Str Str
errToStr = \r ->
when r is
Ok v -> Ok v
Err SomeError -> Err "dummy error"
Err err -> Err (Inspect.toStr err)
I don't have to do anything in the main signature
Brendan Hansknecht (Dec 01 2024 at 18:51):Can you file a bug for this?
Oskar Hahn (Dec 01 2024 at 18:56):Sure: https://github.com/roc-lang/roc/issues/7289
Last updated: Jul 26 2025 at 12:14 UTC