test/fixtures/f2a-witness.sgd

1#lang sigil/doc

2@(import (sigil fs) (sigil process))

3@(define (witness-sentinel name)

4 (write-file-string

5 (string-append (or (getenv "F2A_WITNESS_DIR") "/tmp") "/f2a-witness-" name)

6 "TRIPPED"))

7@(display "F2A-WITNESS-EXPR-RAN\n")

8@(witness-sentinel "expr")

9@(define (tripwire-command . body)

10 ;; An author-defined command. A command at block position receives

11 ;; already-lowered block children, so this returns them rather than

12 ;; re-wrapping them.

13 (display "F2A-WITNESS-COMMAND-RAN\n")

14 (witness-sentinel "command")

15 body)

16@(define (midprose-command . body)

17 ;; POSITION CLASS 5 of 5, the one Fable named. An author command at a

18 ;; NON-BOUNDARY position folds into the enclosing bare-prose run (SD-P3), and

19 ;; no tripwire had ever sat there.

20 (display "F2A-WITNESS-MIDPROSE-COMMAND-RAN\n")

21 (witness-sentinel "midprose-command")

22 ;; `author-normalize` accepts a document node or a proper LIST (which it

23 ;; flattens) and nothing else -- a bare string is refused. An empty list

24 ;; splices to nothing and is legal at every position.

25 '())

26@(define (argument-tripwire-command status-keyword status-value . body)

27 ;; CHANNEL 3. The point of this command is not its body -- it is that its

28 ;; `status:` ARGUMENT is an expression. Reading a field value out of argument

29 ;; syntax is the one way F2a could newly execute author code, and SD-P4's own

30 ;; witness fixture has no argument-position attempt at all, so its

31 ;; stdout-marker witness cannot see this shape by construction.

32 (display "F2A-WITNESS-ARG-COMMAND-RAN\n")

33 body)

34@(define-syntax macro-tripwire

35 (syntax-rules ()

36 ((_ body ...)

37 (begin (display "F2A-WITNESS-MACRO-RAN\n")

38 (witness-sentinel "macro")

39 body ...))))

40@article["F2a no-execution witness" authors: '("F2a")]

42@p{This manuscript exists to prove the no-execution detector can SEE execution.

43Passive inspection must trip none of its channels; evaluating the same file

44through the real @code{#lang} path must trip all of them, and the gate arm runs

45both in the same breath so an all-empty result cannot be mistaken for a clean

46one.}

48@tripwire-command{An author-defined command whose body runs only under

49evaluation.}

51@argument-tripwire-command[status: (begin (display "F2A-WITNESS-ARG-RAN\n")

52 (witness-sentinel "argument")

53 'ready)]{

54@p{The argument above is an ordinary Sigil expression sitting in a keyword

55field. A passive inspector reads it as DATA with @code{syntax->datum}; anything

56that evaluates it instead trips this channel.}

57}

59@(macro-tripwire (display "F2A-WITNESS-MACRO-USE-RAN\n")

60 ;; the macro's expansion must end in a proper list of

61 ;; document nodes, or lowering refuses it AFTER every

62 ;; tripwire has already fired -- which would leave the

63 ;; positive control passing on a run that also errored.

64 '())

66Bare prose, deliberately NOT wrapped in a paragraph command. BOTH ATTEMPTS BELOW

67SIT MID-LINE, and that is the whole point: a command or expression at the START

68of a line is at a BLOCK boundary, where lowering demands it produce document

69nodes. The first version of this fixture put them there and proved only that a

block-position expression runs. Inline expression @(begin (display "F2A-WITNESS-MIDPROSE-RAN\n") (witness-sentinel "midprose") '()) sits here, and inline command @midprose-command{author command} sits here. Under evaluation both fold into

71this paragraph run; in the passive tree both are ordinary flat siblings, which is

72why the passive side sees them at all and why the tripwire has to be HERE rather

73than assumed covered by the block-position one.

75@code["sigil"]{@(display "F2A-WITNESS-CODE-BODY-RAN\n")}

77@end article