[R-pkg-devel] Is this an issue with the package submission?

[Ivan Krylov]

В Thu, 14 Nov 2024 12:27:02 -0500
Shail Choksi <choksishail using gmail.com> пишет:

> The process.txt here:
> https://win-builder.r-project.org/incoming_pretest/ToxicR_24.10.1.4_20241113_200016/specialChecks/clang-san/process.txt
> 
> gives an error. Is this a win-builder issue or a package issue? How
> do I go about debugging this?

Searching for cleancall.c on CRAN yields the 'cli' package:
https://github.com/cran/cli/blob/79876598df96964dee9032362b0773868d624021/src/cleancall.c#L103-L109

The function cb_progress_done from the 'purrr' package, indeed, takes a
different set of argument types:
https://github.com/cran/purrr/blob/de947f6ca0cefdb95f7f0442b08c12c1ef3368ae/src/map.c#L14-L18

This shouldn't break anything on most modern CPU architectures in
practice (R itself is doing similar tricks in order to implement .C()
and .Fortran()), but it is a formal violation of the C standard (and
would crash on certain weird architectures with very large pointers
that encode type information). Since the error is not in your package,
it should be possible to reply to the automatic e-mail (does it say
reply to all or to reply to cran-submission using r-project.org? I don't
remember) and explain the situation.

-- 
Best regards,
Ivan