[R-pkg-devel] parglm: cannot reproduce error

[Benjamin Christoffersen]

Sorry for posting twice. I thought that the mail had not been send out
but it seems like it has been:
https://stat.ethz.ch/pipermail/r-package-devel/2020q1/thread.html

Den man. 16. mar. 2020 kl. 12.19 skrev Benjamin Christoffersen
<boennecd using gmail.com>:
>
> Dear Sirs,
>
> Recently, my parglm package was pulled of CRAN because of a segfault
> on Solaris and an issue on the OpenBLAS additional issues test.
>
> I started by trying to reproduce the Solaris issue but failed to
> reproduce the error
> with `rhub::checkonsolaris()`. Then I tried to reproduce the OpenBLAS error.
> First, I build R-devel with
> `–with-blas="/usr/lib/x86_64-linux-gnu/openblas/libblas.so.3"` and
> `–with-lapack="/usr/lib/x86_64-linux-gnu/openblas/liblapack.so.3"`
> which one gets from `libopenblas-dev` if I recall correctly on my own
> Ubuntu machine (version 18.04, gcc 8.3.0). However, there where no
> issues or segfaults when I ran `CMD check –as-cran
> –no-stop-on-test-error`.
>
> Then I build the following docker image:
> https://gist.github.com/boennecd/fd0625daf11aae3304e9e9a9a6cfea7d
>
> where I do as described in the following two links (I think):
> https://www.stats.ox.ac.uk/pub/bdr/Rblas/README.txt &
> https://www.stats.ox.ac.uk/pub/bdr/Rconfig/r-devel-linux-x8664-fedora-gcc
>
> However when I started the container and ran:
>  - `/opt/R-devel/lib64/R/bin/R CMD check –as-cran
> –no-stop-on-test-error –use-valgrind`
>  - `/opt/R-devel/lib64/R/bin/R CMD check –as-cran –no-stop-on-test-error`
>
> then I got no issues or segfaults.
>
> So, although that other people have asked similar questions before,
> can anyone spot what I did wrong when I tried to reproduce the error
> and if not what should I do? Any help is much appreciated.
>
> There are no other issues with both ASAN and UBSAN checks with gcc and
> clang on my own machine.
>
> Sincerely Yours,
> Benjamin Christoffersen