Re: master: Make *READER-PACKAGE* be more nearly like binding *PACKAGE*

::: UNEXPECTED-FAILURE :READ-IN-PACKAGE-SYNTAX due to SIMPLE-ERROR:
"The assertion
(EQ :VIOLATED!
(HANDLER-CASE (READ-FROM-STRING "cl::'foo")
(PACKAGE-LOCK-VIOLATION NIL :VIOLATED!)))
failed with
(HANDLER-CASE (READ-FROM-STRING "cl::'foo")
(PACKAGE-LOCK-VIOLATION NIL :VIOLATED!))
= 'COMMON-LISP::FOO."

On Mon, Nov 17, 2025 at 9:50 PM snuglas via Sbcl-commits
<[email protected]> wrote:
>
> The branch "master" has been updated in SBCL:
>        via  ce22877b3b6cbdb9816016c44c73e7a575ab3c4e (commit)
>       from  1a339b8f171488b492b2190e502b16924688a6c5 (commit)
>
> - Log -----------------------------------------------------------------
> commit ce22877b3b6cbdb9816016c44c73e7a575ab3c4e
> Author: Douglas Katzman <[email protected]>
> Date:   Mon Nov 17 13:07:19 2025 -0500
>
>     Make *READER-PACKAGE* be more nearly like binding *PACKAGE*
>
>     Regarding package locks, it needs to be the same as being in *PACKAGE* as far as
>     INTERN is concerned, otherwise it's not a very useful shortcut.
>     Binding *PACKAGE* could have surprising consequences, as could burying more logic
>     into PACKAGE-LOCK-VIOLATION-P. As it happens, %INTERN has a lock bypass mode.
> ---
>  src/code/reader.lisp            | 4 +++-
>  tests/package-locks.impure.lisp | 9 ++++++++-
>  2 files changed, 11 insertions(+), 2 deletions(-)
>
> diff --git a/src/code/reader.lisp b/src/code/reader.lisp
> index b72e86b1a..bda70c126 100644
> --- a/src/code/reader.lisp
> +++ b/src/code/reader.lisp
> @@ -1602,7 +1602,9 @@ extended <package-name>::<form-in-package> syntax."
>                                   (if (token-buf-only-base-chars buf)
>                                       (%readtable-symbol-preference rt)
>                                       'character)
> -                                 nil)))))))))
> +                                 ;; reader-package behaves as if *package* were that package.
> +                                 ;; Hence it should be allowed to create new symbols.
> +                                 (eq pkg *reader-package*))))))))))
>
>  ;;; For semi-external use: Return 3 values: the token-buf,
>  ;;; a flag for whether there was an escape char, and the position of
> diff --git a/tests/package-locks.impure.lisp b/tests/package-locks.impure.lisp
> index 551860e22..b71f9ce74 100644
> --- a/tests/package-locks.impure.lisp
> +++ b/tests/package-locks.impure.lisp
> @@ -677,4 +677,11 @@
>       (progv vars vals))
>     (('(test:*special*) nil) (condition 'symbol-package-locked-error))))
>
> -
> +(with-test (:name :reader-package)
> +  (lock-package (make-package "SOMEPACKAGE"))
> +  (let ((form (read-from-string "somepackage::(cl:defvar myvar 3)")))
> +    (assert (eq (symbol-package (second form)) (find-package "SOMEPACKAGE"))))
> +  ;; I don't see why deleting a package can't be done if locked. Locking should concern
> +  ;; the contents of the package, not what packages are in my system. Anyway, different problem.
> +  (unlock-package "SOMEPACKAGE")
> +  (delete-package "SOMEPACKAGE"))
>
> -----------------------------------------------------------------------
>
>
> hooks/post-receive
> --
> SBCL
>
>
> _______________________________________________
> Sbcl-commits mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/sbcl-commits


_______________________________________________
Sbcl-commits mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/sbcl-commits