On 7 January 2019 at 12:29, Sébastien Villemot wrote: | Dear Dirk and Jörg-Volker, | | On Mon, 7 Jan 2019 08:54:22 +0100 Jörg-Volker Peetz <jvpe...@web.de> wrote: | | > the problem I see, shows when inspecting /usr/lib/R/modules/lapack.so with | > | > $ ldd /usr/lib/R/modules/lapack.so | grep blas | > libblas.so.3 => /usr/lib/x86_64-linux-gnu/libblas.so.3 (0x00007f58c61b5000) | > libopenblas.so.0 => /usr/lib/x86_64-linux-gnu/libopenblas.so.0 (0x00007f58c2dbb000) | > | > Thus, lapack.so of package r-base-core is linked to libopenblas.so.0 | > which means that r-base-core won't work without package libopenblas-base. | > And that is not intended AFAIU. | | As explained in <1546860281.3311.22.ca...@debian.org>, I think there is | no bug. | | ldd lists libopenblas.so.0 as a dependency, but it is not actually | needed. I guess it's only listed because it's a *indirect* dependency | through libblas.so.3, when openblas is selected as the BLAS | alternative. | | As a matter of fact, R runs fine without libopenblas-base installed.
Thank you. I had not seen your "in-flight" message when I replied here and on debian-science. We agree. And there may well be something "unusual to the untrained eye" but this has worked since way pretty much since the 1990s thanks to some really clever and novel work done mostly by Camm Maguire. We still take advantage of it now. Dirk -- http://dirk.eddelbuettel.com | @eddelbuettel | e...@debian.org
