bug#57573: disarchive: Test "[prop] Writing is reversible" fails

bug#57573: disarchive: Test "[prop] Writing is reversible" fails

[Maxime Devos]

X-Debbugs-CC: Timothy Sample <samplet@ngyro.com>

Hi,

disarchive fails to build (on core-updates, more precisely as part of 
the antiox jobset on ci.guix.gnu.org).

The failing test is:

> test-name: [prop] Writing is reversible
> location: tests/kinds/octal.scm:128
> source:
> + (test-assert
> +   "[prop] Writing is reversible"
> +   (quickcheck
> +     (property
> +       ((octal $octal))
> +       (test-when
> +         (valid-octal? octal)
> +         (begin
> +           (equal?
> +             octal
> +             (decode-octal (encode-octal octal))))))))
> Falsifiable after 25 tests.
> Seed: 318171092
> octal = #<<unstructured-octal> value: 0 source: #<<zero-string> value: 
> "0" trailer: "?">>
> actual-value: #f
> result: FAIL
(This appears to be a different test failure than in 
<https://issues.guix.gnu.org/48114>)

TBI ...

Greetings,
Maxime.

bug#57573: disarchive: Test "[prop] Writing is reversible" fails

[Maxime Devos]

Here's a reproducer:

(define o (make-unstructured-octal 0 (make-zero-string "0" #\䤄)))
(test-equal "Writing is reversible (#57573)"
         o (decode-octal (encode-octal octal)))

bug#57573: disarchive: Test "[prop] Writing is reversible" fails

[Maxime Devos]

On 04-09-2022 10:24, Maxime Devos wrote:
> Here's a reproducer:
>
> (define o (make-unstructured-octal 0 (make-zero-string "0" #\䤄)))
> (test-equal "Writing is reversible (#57573)"
>         o (decode-octal (encode-octal octal)))
>
That was bogus (octal is undefined), try

>
> (define o (make-unstructured-octal 0 (make-zero-string "0" "䤄")))
> (unless (valid-octal? o) (error "oops"))
> (test-equal "Writing is reversible (#57573)"
>         o (decode-octal (encode-octal o)))
instead. Error message:
> test-name: Writing is reversible (#57573)
> location: tests/kinds/octal.scm:130
> source:
> + (test-equal
> +   "Writing is reversible (#57573)"
> +   o
> +   (decode-octal (encode-octal o)))
> expected-value: #<<unstructured-octal> value: 0 source: 
> #<<zero-string> value: "0" trailer: "䤄">>
> actual-value: #<<padded-octal> value: 0 width: 1 padding: #\0 trailer: 
> "䤄">
> result: FAIL
Greetings,
Maxime.

bug#57573: disarchive: Test "[prop] Writing is reversible" fails

[Timothy Sample]

Hi Maxime,

Maxime Devos <maxime.devos@student.kuleuven.be> writes:

> (define o (make-unstructured-octal 0 (make-zero-string "0" "䤄")))
> (unless (valid-octal? o) (error "oops"))
> (test-equal "Writing is reversible (#57573)"
>         o (decode-octal (encode-octal o)))

Right.  That unstructured octal should be invalid.  By the definition in
the code, it has too much “structure”.  It’s a test issue, the actual
code is fine.  I’ll patch the tests and make a point release in the
coming days.

Thanks for the heads up and the analysis, and sorry for the
inconvenience!


-- Tim

P.S. I’m excited about antioxidant – good work and happy hacking!