bug#66042: 30.0.50; ERT test fails while trying to indent pretty-print

bug#66042: 30.0.50; ERT test fails while trying to indent pretty-print

[Protesilaos Stavrou]

Dear maintainers,

I have encountered a possible bug with an ERT test where it is trying to
perform indentation and fails with:

    lisp-indent-initial-state: Wrong type argument: wholenump, -2

This is the test I am running:

    (defun test-case (text pos expected)
      (with-temp-buffer
        (insert text)
        (goto-char pos)
        (backward-kill-word 1)
        (should (string-equal (buffer-string) expected))))

    (ert-deftest my-test ()
      (test-case "aaa)))" 7 "aaa"))

Upon investigating further, I found that 'ert--pp-with-indentation-and-newline'
performs an 'indent-sexp'.  When I comment that out and redefine the
function, the test no longer reports an indentation error with
wholenump.  The test fails gracefully, as expected.  Thus, the whole
file I am using is this:

    (require 'ert)

    (defun ert--pp-with-indentation-and-newline (object)
      "Pretty-print OBJECT, indenting it to the current column of point.
    Ensures a final newline is inserted."
      (let ((begin (point))
            (pp-escape-newlines t)
            (print-escape-control-characters t))
        (pp object (current-buffer))
        (unless (bolp) (insert "\n"))
        (save-excursion
          (goto-char begin)
          ;; (indent-sexp)
          )))

    (defun test-case (text pos expected)
      (with-temp-buffer
        (insert text)
        (goto-char pos)
        (backward-kill-word 1)
        (should (string-equal (buffer-string) expected))))

    (ert-deftest my-test ()
      (test-case "aaa)))" 7 "aaa"))

To be clear, my test should fail, but that is expected.  I encountered
the ERT wholenump issue which prevented me from going further.

Is the indentation expected behaviour?  Can we disable any/all
indentation of sexps within a test?

Thank you for your time,
Protesilaos (or simply "Prot")

-- 
Protesilaos Stavrou
https://protesilaos.com

bug#66042: 30.0.50; ERT test fails while trying to indent pretty-print

[Mauro Aranda]

Hi Prot,

Protesilaos Stavrou <info@protesilaos.com> writes:

 > Dear maintainers,
 >
 > I have encountered a possible bug with an ERT test where it is trying to
 > perform indentation and fails with:
 >
 >     lisp-indent-initial-state: Wrong type argument: wholenump, -2
 >
 > This is the test I am running:
 >
 >     (defun test-case (text pos expected)
 >       (with-temp-buffer
 >         (insert text)
 >         (goto-char pos)
 >         (backward-kill-word 1)
 >         (should (string-equal (buffer-string) expected))))
 >
 >     (ert-deftest my-test ()
 >       (test-case "aaa)))" 7 "aaa"))

I tried to reproduce it on current master, but I couldn't.  I ran the
test interactively, in case that makes a difference.

bug#66042: 30.0.50; ERT test fails while trying to indent pretty-print

[Stefan Kangas]

Protesilaos Stavrou <info@protesilaos.com> writes:

> I have encountered a possible bug with an ERT test where it is trying to
> perform indentation and fails with:
>
>     lisp-indent-initial-state: Wrong type argument: wholenump, -2
>
> This is the test I am running:
>
>     (defun test-case (text pos expected)
>       (with-temp-buffer
>         (insert text)
>         (goto-char pos)
>         (backward-kill-word 1)
>         (should (string-equal (buffer-string) expected))))
>
>     (ert-deftest my-test ()
>       (test-case "aaa)))" 7 "aaa"))

I can't reproduce this on current master.  Evaluating the above two
forms in emacs -Q and then saying

    M-x ert RET t RET

yields a new buffer "*ert*" containing

    Selector: t
    Passed:  0
    Failed:  1 (1 unexpected)
    Skipped: 0
    Total:   1/1

    Started at:   2023-09-17 13:44:28+0200
    Finished.
    Finished at:  2023-09-17 13:44:28+0200

    F

    F my-test
        (ert-test-failed
         ((should (string-equal (buffer-string) expected)) :form
          (string-equal "" "aaa") :value nil :explanation
          (arrays-of-different-length 0 3 "" "aaa" first-mismatch-at 0)))

Which version of Emacs are you running?

Can you reproduce it in emacs -Q?

bug#66042: 30.0.50; ERT test fails while trying to indent pretty-print

[Protesilaos Stavrou]

[-- Attachment #1: Type: text/plain, Size: 1483 bytes --]

Hello Stefan, Mauro,

> From: Stefan Kangas <stefankangas@gmail.com>
> Date: Sun, 17 Sep 2023 04:47:29 -0700

> [... 20 lines elided]

> I can't reproduce this on current master.  Evaluating the above two
> forms in emacs -Q and then saying

> [... 23 lines elided]

> Which version of Emacs are you running?
>
> Can you reproduce it in emacs -Q?

Yes, I can reproduce it with 'emacs -Q' on Emacs trunk.  I tried it just
now with the following code.  See attached screenshot.

    (require 'ert)

    (defun ert--pp-with-indentation-and-newline (object)
      "Pretty-print OBJECT, indenting it to the current column of point.
    Ensures a final newline is inserted."
      (let ((begin (point))
            (pp-escape-newlines t)
            (print-escape-control-characters t))
        (pp object (current-buffer))
        (unless (bolp) (insert "\n"))
        (save-excursion
          (goto-char begin)
          (indent-sexp)
          )))

    (ert-deftest my-kill-backward-test-close-parens ()
      "aaa)))"
      (test-case "aaa)))" 7 "aaa"))

    (defun test-case (text pos expected)
      (with-temp-buffer
        (insert text)
        (goto-char pos)
        (backward-kill-word 1)
        (should (string-equal (buffer-string) expected))))

If I restart Emacs, comment out (indent-sexp), then evaluate the buffer,
and then the ERT test, everything works as expected.  There is no
'wholenump' error.

All the best,
Prot

-- 
Protesilaos Stavrou
https://protesilaos.com

[-- Attachment #2: 2023-09-18_07:50:02_1274x1056.png --]
[-- Type: image/png, Size: 35248 bytes --]

bug#66042: 30.0.50; ERT test fails while trying to indent pretty-print

[Mauro Aranda]

On 18/9/23 01:54, Protesilaos Stavrou wrote:
 > Hello Stefan, Mauro,
 >
 >> From: Stefan Kangas <stefankangas@gmail.com>
 >> Date: Sun, 17 Sep 2023 04:47:29 -0700
 >
 >> [... 20 lines elided]
 >
 >> I can't reproduce this on current master.  Evaluating the above two
 >> forms in emacs -Q and then saying
 >
 >> [... 23 lines elided]
 >
 >> Which version of Emacs are you running?
 >>
 >> Can you reproduce it in emacs -Q?
 >
 > Yes, I can reproduce it with 'emacs -Q' on Emacs trunk.  I tried it just
 > now with the following code.  See attached screenshot.
 >
 >      (require 'ert)
 >
 >      (defun ert--pp-with-indentation-and-newline (object)
 >        "Pretty-print OBJECT, indenting it to the current column of point.
 >      Ensures a final newline is inserted."
 >        (let ((begin (point))
 >              (pp-escape-newlines t)
 >              (print-escape-control-characters t))
 >          (pp object (current-buffer))
 >          (unless (bolp) (insert "\n"))
 >          (save-excursion
 >            (goto-char begin)
 >            (indent-sexp)
 >            )))
 >
 >      (ert-deftest my-kill-backward-test-close-parens ()
 >        "aaa)))"
 >        (test-case "aaa)))" 7 "aaa"))
 >
 >      (defun test-case (text pos expected)
 >        (with-temp-buffer
 >          (insert text)
 >          (goto-char pos)
 >          (backward-kill-word 1)
 >          (should (string-equal (buffer-string) expected))))
 >

I can reproduce it now.  Your original recipe didn't include a docstring
for the ERT test.