[R-pkg-devel] UBSAN Problems with Memory Sanitizer

Maximilian Leodolter m@ximili@n@leodolter @ending from gm@il@com
Wed Jan 2 22:31:04 CET 2019


Ok thanks Dirk! I'll try to explain to the CRAN team.
Yes the vignette-issue because of duplicated file titles is solved.

Best, Max

On Wed, Jan 2, 2019 at 5:02 PM Dirk Eddelbuettel <edd using debian.org> wrote:

>
> On 2 January 2019 at 16:17, Maximilian Leodolter wrote:
> | I just wanted to upload a new version of my package (IncDTW) but got the
> | message that my old version causes the following error, even though it
> | works perfectly fine.
> |
> | The error message says:
> |
> | runtime error: member call on address 0x7f4c77dcb200 which does not
> | point to an object of type 'scheduler'
> | 0x7f4c77dcb200: note: object is of type
> | 'tbb::internal::custom_scheduler<tbb::internal::IntelSchedulerTraits>'
> |
> |
> | The error messages are given here in detail
> |
> https://www.stats.ox.ac.uk/pub/bdr/memtests/clang-UBSAN/IncDTW/tests/testthat.Rout
> | and here
> |
> |
> https://www.stats.ox.ac.uk/pub/bdr/memtests/gcc-UBSAN/IncDTW/tests/testthat.Rout
> |
> |
> | And this is the cran-github mirror of my package
> | https://github.com/cran/IncDTW
> |
> | I have no idea where the error messages point to and what causes the
> error,
> | even less how to solve it.
> |
> |  I'm new to C++ and RcppParallel and don't really know what to do. Can
> | someone please help?
>
> Rock, meet hard place. It seems you are getting an UBSAN error from an
> external library. That can happen, and sometimes you get to explain it to
> CRAN so that the package gets admitted.
>
> Having glanced at the output below the page at
>   https://cloud.r-project.org/web/checks/check_results_IncDTW.html
> I think there is little you can do on your end about the UBSAN issues.
>
> I presume you did your part to make the NOTE from the vignette disappear?
>
> Dirk
>
> --
> http://dirk.eddelbuettel.com | @eddelbuettel | edd using debian.org
>

	[[alternative HTML version deleted]]



More information about the R-package-devel mailing list