structure(<primitive function>, ...) is sticky: a bug, or should it be an error?

Wed, Apr 2, 2025 8:24 PM

Thank you Martin - I've been following your commits to R-devel and
R-beta. FWIW, I've just verified that it works as expected - I scanned
the base environment for all primitive functions and then verified
that attr(x, "abc") <- TRUE gives an error on R-devel;

$ R --vanilla
R Under development (unstable) (2025-04-02 r88098) -- "Unsuffered Consequences"

Error in attr(sum, "abc") <- TRUE : cannot set attribute on a 'builtin'

a warning on R-beta;

$ R --vanilla

Warning message:
In attr(sum, "abc") <- TRUE :
  Setting attributes on primitive functions is deprecated and will be disabled

and neither on R release.

For the warning in R 4.5.0 beta, have you considered making it a
warning of class 'deprecatedWarning'? Not a big deal, but wanted to
mention it, just in case.

/Henrik

On Fri, Mar 21, 2025 at 10:07?AM Henrik Bengtsson

<henrik.bengtsson at gmail.com> wrote:

On Fri, Mar 21, 2025 at 8:55?AM Martin Maechler
<maechler at stat.math.ethz.ch> wrote:

Martin Maechler
    on Fri, 21 Mar 2025 14:49:16 +0100 writes:

Henrik Bengtsson
    on Thu, 20 Mar 2025 11:55:05 -0700 writes:

    >>> I'm pretty convinced we should fix it by checking for
    >>> primitive functions inside the C code of `attributes<-` :
    >>> arguably the bug is really there, rather than in
    >>> structure().
    >>>
    >>> Patches are welcome (via R's Bugzilla or just here).

    >> Thank you Martin. I'll make sure I create a brief BugZilla
    >> report on this, and hopefully a follow with a patch later
    >> on.

    >> One question on urgency or not: Is it too late to get such
    >> a change in for the R 4.5.0 release? I suspect so, because
    >> it has a potential of breaking existing packages. But if
    >> there's a possibility of fixing this in R 4.5.0, I'll make
    >> this a top priority.  Please let me know.

    >> /Henrik

    > In such cases, I always go to  https://developer.r-project.org/
    > and look at the release schedule.

    > Important for this Q seems

    > * Friday 2025-03-14: GRAND-FEATURE FREEZE (4.5.0 alpha)
    > * Friday 2025-03-28: FEATURE FREEZE (4.5.0 beta)
    > * Friday 2025-04-04: CODE FREEZE (4.5.0 RC)

    > So, yes there's a possibility  ... but as you mentioned, too,
    > it depends quite a bit on if the fix is easily affectiong useR
    > and CRAN/Bioc package space.

    > Martin

As a matter of fact, I'm testing already a simple patch I wrote
myself.
So, sorry, but I must renounce the above  "Patches are welcome".

Excellent and even better (for me). Thank you Martin!

/Henrik

Martin