Better error message when using connection from wrong process or after closing (Erlang)

Hello hello!

I accidentally introduced a bug in my application where I was passing a connection to another erlang actor and it kept crashing. Took me a while to figure out that I was passing the connection from with_connection after the scope was already gone. Obviously on me, but I was wondering if we can provide a better error message for this. Same for when you close the connection and use it afterwards.

I am not super familiar with the details of the erlang implementation, so maybe this is not even possible. Happy to make a PR though if this is deemed viable.

Example Code:

import app/env
import gleam/dynamic/decode
import gleam/erlang/process
import sqlight

pub fn foo(conn: sqlight.Connection) {
  echo sqlight.query(
    "select 1",
    on: conn,
    with: [],
    expecting: decode.success(1),
  )
}

pub fn bar() {
  let env = env.get_env()
  use conn <- sqlight.with_connection(env.db_path)
  process.spawn(fn() { foo(conn) })
}

pub fn main() {
  let _ = bar()
  process.sleep_forever()
}

Output:

gleam dev
  Compiling app
   Compiled in 0.41s
    Running app_dev.main
=CRASH REPORT==== 29-Oct-2025::12:07:16.917653 ===
  crasher:
    initial call: app_dev:'-bar/0-anonymous-0-'/0
    pid: <0.115.0>
    registered_name: []
    exception error: bad argument
      in function  esqlite3_nif:error_info/1
         called as esqlite3_nif:error_info(#Ref<0.302291343.3268280346.105147>)
      in call from sqlight_ffi:to_error/2 (/app/build/dev/erlang/sqlight/_gleam_artefacts/sqlight_ffi.erl:61)
      in call from sqlight:query/4 (src/sqlight.gleam:387)
      in call from app_dev:foo/1 (dev/app_dev.gleam:11)
    ancestors: [<0.82.0>]
    message_queue_len: 0
    messages: []
    links: [<0.82.0>]
    dictionary: []
    trap_exit: false
    status: running
    heap_size: 233
    stack_size: 29
    reductions: 100
  neighbours:
    neighbour:
      pid: <0.82.0>
      registered_name: []
      initial_call: {erlang,apply,2}
      current_function: {code_server,call,1}
      ancestors: []
      message_queue_len: 0
      links: [<0.10.0>,<0.115.0>]
      trap_exit: false
      status: waiting
      heap_size: 233
      stack_size: 16
      reductions: 3909
      current_stacktrace: [{code_server,call,1,[{file,"code_server.erl"},{line,163}]},
                  {code,ensure_loaded,1,[{file,"code.erl"},{line,578}]},
                  {error_handler,undefined_function,3,
                      [{file,"error_handler.erl"},{line,86}]},
                  {gleam_erlang_ffi,sleep_forever,0,
                      [{file,
                           "/app/build/dev/erlang/gleam_erlang/_gleam_artefacts/gleam_erlang_ffi.erl"},
                       {line,27}]},
                  {app@@main,run_module,1,
                      [{file,
                           "/app/build/dev/erlang/app/_gleam_artefacts/app@@main.erl"},
                       {line,24}]}]
runtime error: Erlang exit

An error occurred outside of Gleam.

erlang:exit(Badarg)

stacktrace:
  esqlite3_nif.error_info unknown source
  sqlight_ffi.to_error /app/build/dev/erlang/sqlight/_gleam_artefacts/sqlight_ffi.erl:61
  sqlight.query src/sqlight.gleam:387
  app_dev.foo dev/app_dev.gleam:11
  proc_lib.init_p proc_lib.erl:317

Start at sqlight_ffi.erl line 61, where the reported bad argument occurs, and trace how src/sqlight.gleam line 387 reaches it. Reproduce the provided cross-process example and the closed-connection case; done means both produce a useful connection error instead of the shown crash.