panic when trying to asserta/1 a process handle

[danilp-id]

?- use_module(library(process)).
   true.
?- process_create("true", [], [process(P)]), asserta(process(P)).


thread 'main' (115132) panicked at src/machine/loader.rs:1416:115:
called `Result::unwrap()` on an `Err` value: ()
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

[danilp-id]

The same thing happens if instead of asserta you do write(P) or write(foo(P)).

[Skgland]

The error points to this line in MachineState::read_term_from_heap

scryer-prolog/src/machine/loader.rs

Line 1416 in 974722b

term_stack.push(Term::Literal(Cell::default(), Literal::try_from((addr, &self.arena.f64_tbl)).unwrap()));

we shouldn't be using unwrap there.

Literal::try_from doesn't handle Cons heap cells that aren't Integer or Rational

scryer-prolog/src/types.rs

Lines 340 to 352 in 974722b

	(HeapCellValueTag::Cons, cons_ptr) => {
	match_untyped_arena_ptr!(cons_ptr,
	(ArenaHeaderTag::Integer, n) => {
	Ok(Literal::Integer(n))
	}
	(ArenaHeaderTag::Rational, n) => {
	Ok(Literal::Rational(n))
	}
	_ => {
	Err(())
	}
	)
	}

I am also a bit surprised to see that this doesn't at least have a debug assert to check that this ends up with the correct tag. We appear to assume that the high/low bits (depends on endianness) that end up being the tag are always zero.
Is it even correct for us to use native endianness? This means it depends on the endianness whether the high/low bits end up in the tag bits, but setting alignment on the pointee type would only ensure that the low bits are zero.

scryer-prolog/src/types.rs

Lines 588 to 590 in 974722b

	pub fn from_ptr_addr(ptr_bytes: usize) -> Self {
	HeapCellValue::from_bytes((ptr_bytes as u64).to_ne_bytes())
	}

scryer-prolog/src/arena.rs

Lines 83 to 85 in 974722b

	#[repr(align(8))]
	#[derive(Copy, Clone, Debug)]
	pub struct ArenaHeader {

[Skgland]

Does the same happen for streams as well? (I would expect so)?
What would the expected behavior (besides not crashing) be?

• asserta I would expect to work with equality based on the arena object address?
• write should probably print it as $proccess(1234) where 1234 is the process id
  to match the behavior at

  scryer-prolog/src/heap_print.rs

  Lines 1769 to 1784 in 974722b

  	(ArenaHeaderTag::ChildProcess, process) => {
  	let process_atom = atom!("$process");
  	if self.format_struct(max_depth, 1, process_atom) {
  	let atom = TokenOrRedirect::NumberFocus(max_depth, NumberFocus::Unfocused(Number::Fixnum(Fixnum::build_with(process.id()))), op);
  	let process_root = self.state_stack.pop().unwrap();
  	self.state_stack.pop();
  	self.state_stack.pop();
  	self.state_stack.push(atom);
  	self.state_stack.push(TokenOrRedirect::Open);
  	self.state_stack.push(process_root);
  	}

[Skgland]

Does the same happen for streams as well? (I would expect so)?

Quick tests shows that for streams write appears to work while asserta panics with the same error.

> scryer-prolog.exe --version
cargo:0.10.0

> scryer-prolog.exe
?- current_output(S), asserta(stream(S)).

thread 'main' panicked at C:\Users\<username>\.cargo\registry\src\index.crates.io-1949cf8c6b5b557f\scryer-prolog-0.10.0\src\machine\loader.rs:1416:115:
called `Result::unwrap()` on an `Err` value: ()
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

> scryer-prolog.exe
?- current_output(S), write(stream(S)).
stream(user_output)   S = user_output.
?-

[notoria]

This is related to #1855.

?- current_input(S), functor(S, N, A), S == N.
   S = user_input, N = user_input, A = 0.
?- current_input(S), functor(S, N, A), S == N, atom(N).
   false. % unexpected.
?- current_input(S), functor(S, N, A), S == N, assertz(stream(N)).

thread 'main' (3452853) panicked at src/machine/loader.rs:1416:115:
called `Result::unwrap()` on an `Err` value: ()
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

[danilp-id]

Actually, I don't seem to be able to reproduce the write/1 issue.

Aside from that, bug also happens with sockets and request handles:

?- use_module(library(sockets)).
   true.
?- socket_server_open('0.0.0.0':5234, Socket), asserta(mysock(Socket)).

thread 'main' (9504) panicked at src/machine/loader.rs:1416:115:
called `Result::unwrap()` on an `Err` value: ()
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

?- use_module(library(http/http_server)).
   true.
?- [user].
foo(Req,Res) :- asserta(foo(Req)).
?- http_listen(4314, [get(/, foo)]).
Listening at 0.0.0.0:4314
2026-04-21 (16:27:58) get /

thread 'main' panicked at src/machine/loader.rs:1416:115:
called `Result::unwrap()` on an `Err` value: ()
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

[danilp-id]

Interestingly this doesn't seem to be an issue with bb_put/2:

?- use_module(library(sockets)).
   true.
?- use_module(library(iso_ext)).
   true.
?- socket_server_open('0.0.0.0':5234, Socket), bb_put(mysock, Socket).
   Socket = '$tcp_listener'('0.0.0.0',5234).
?- bb_get(mysock, Sock).
   Sock = '$tcp_listener'('0.0.0.0',5234).

?- use_module(library(iso_ext)).
   true.
?- use_module(library(process)).
   true.
?- process_create("true", [], [process(P)]), bb_put(proc, P).
   P = '$process'(31206).
?- bb_get(proc, P).
   P = '$process'(31206).
?- bb_get(proc, P), process_wait(P, R).
   P = '$process'(31206), R = exit(0).