Currently, if more than one attribute is missing or wrong, which attribute is reported as problematic depends on the order the attribute shows up in the attribute pair list.
> Recent bug fixes in the development version of R,
> "R-devel", (see the short thread starting here:
> https://stat.ethz.ch/pipermail/r-devel/2018-March/075644.html
> ) have lead your packages to no longer pass 'R CMD
> check' with that version of R-devel:
> https://www.r-project.org/web/checks/check_results_nima.html
> https://www.r-project.org/web/checks/check_results_vetr.html
> In both cases, the reason is a *too strict* test, you not
> only check that the attributes (of a certain data frame)
> are correct, but you *wrongly* also check that the list
> which attributes(<df>) returns has the attributes in a
> given order.
> Instead please fix your check e.g., to use identical() if
> appropriate, because even identical() with its default
> argument 'attr.as.set = TRUE' does *not* distinguish
> objects where only the attributes are re-ordered.
> Best regards,
> Martin Maechler ETH Zurich and R Core Team
Currently, if more than one attribute is missing or wrong, which attribute is reported as problematic depends on the order the attribute shows up in the attribute pair list.