HELP INCOMPLETE_CODE                    Refuse source that would wedge the itemiser

    uses incomplete_code;

incomplete_code(STRING) -> WHY
    false when STRING is structurally complete, otherwise a short
    description: 'unterminated string', 'unclosed bracket',
    'unclosed /* comment'.

The compiler recovers cleanly from a mishap INSIDE a complete stream:
the interrupt trap fires, the stack is repaired, and the session
carries on.  A stream that ENDS mid-token is different — an unclosed
string, bracket or comment leaves shared itemiser state that no trap
can repair, and every later chunk is read as part of the unfinished
one.

So any server that feeds a live session code from elsewhere — an agent
over MCP, an editor over LSP, a REPL over a socket — has to refuse that
case up front:

    uses incomplete_code;
    lvars why = incomplete_code(code);
    if why then
        'refused: ' <> why
    else
        pop11_compile(stringin(code))
    endif;

Note what is deliberately NOT checked.  An unfinished `define' is
fine: it leaves the compiler waiting for the rest, which is exactly how
one builds a procedure interactively.  The check is lexical only — it
knows about `;;;' comments, nesting `/* */' comments, `'strings'`,
`"words' (whose closing quote is optional), `c` character literals and
the three bracket pairs.  It is not a parser and does not pretend to
be one.

See also: LIB * JSONRPC, pop/mcp/pop11_mcp.p, pop/lsp/pop11_lsp.p,
tools/tests/test_jsonrpc.p.

--- pop/help/incomplete_code