Function definition is ignored

[ytoml]

Describe the bug
Function declaration ignored, i.e. like encountering NameError on f(1) after f x = x.

Reproducible code

f x = x
f
# compiler complains even f is defined.
Error[#0213]: File <stdin>, line 1, in <module>
1│ f(1)
   ^
NameError: f is not defined

Expected behavior
Call f.

Additional context
Actually I found it's originated in Desugarer::desugar_pattern.
Its matching ignores Expr::Def(Def { sig: Signature::Subr(v)..}) like:

while let Some(chunk) = module.lpop() {
    match chunk {
        Expr::Def(def) => {
            if let Signature::Var(v) = def.sig {
                /* desugaring */
            } // if end
        },
        other => new.push(other),
    } // match end
} 

Thus, function parsed as Subr vanishes there.

However just fixing this as

match chunk {
    Expr::Def(Def { sig: Signature::Var(v) .. }) => { /* omit */ }
    other => /* omit */
}

produces stack overflow (eternal recursion) in compile process.
Maybe compiler also have some bug as the eternal recursion is not desireble if it's fatal error, but we have to add desugaring logic for Signature::Subr here first, rather than just passing it through this matching, right?

If so, I'm willing to implement them as it seems to be good first implementation to get used to Erg's syntax.
(And maybe fixing compiler bug is also not difficult and I'm willing to do it too.)

[ytoml]

Maybe compiler also have some bug as the eternal recursion is not desireble if it's fatal error, but we have to add desugaring logic for Signature::Subr here first, rather than just passing it through this matching, right?

Maybe I was wrong because desugar_pattern seems (by its name) to be just for desugaring pattern matching statements but not functions or procedures.

[ytoml]

Now I'm determined to keep this pattern matching as it is except for ignoring Subr, and look into compiler to debug eternal recursion.

[mtshiba]

You are correct, this bug is due to a leak in the pattern match. As far as I can see in the debug mode, your fix solves this issue, and the stack overflow seems to be a type inference bug, not a parser bug.

In the meantime, could you please send me a PR that fixes this bug?

[mtshiba]

I fixed the infinite recursion issue in #96.

[ytoml]

Thank you for information! I will open PR soon.