On 25 November 2018 at 21:59, Paul Gevers wrote: | Hi Dirk, | | On 25-11-18 21:42, Dirk Eddelbuettel wrote: | > The new version of readr relies on rpath to encode a path. We are caught | > between a rock and hard place: as shipped (and built my yesterday), the | > package "builds" but does not longer autotest you saw here. | | I have difficulty parsing what you mean here. I guess you mean | r-cran-readr doesn't have an autopkgtest by itself anymore?
No I said you may want to consider turning the autotest off by choice while the upstream code remains broken (in the "simultaneously requiring two different rpath settings" sense). | > Not quite sure what we can do -- maybe for once suspend autotests on this one? | | Are you talking about the build-time tests? Because aren't the | autopkgtests of r-cran-alakazam and r-cran-tidyverse saying that the | current package of r-cran-readr in unstable is broken? Of course. And that is what you reported to me, the maintainer of r-cran-readr. And I am trying to explain to you, and failing, that this is a bad upstream choice I have limited influence over which caused this. Dirk -- http://dirk.eddelbuettel.com | @eddelbuettel | [email protected]
