On Thu, Oct 18, 2018 at 06:34:23AM -0500, Dirk Eddelbuettel wrote: > > | diff --git a/debian/control b/debian/control > | index a6114be..a98b487 100644 > | --- a/debian/control > | +++ b/debian/control > | @@ -11,7 +11,7 @@ Homepage: https://cran.r-project.org/package=BH > | Package: r-cran-bh > | Architecture: all > | Multi-Arch: foreign > | -Depends: ${misc:Depends}, libboost-dev > | +Depends: ${misc:Depends}, libboost1.67-dev > | Description: (Virtual) GNU R package to provide BH > | The CRAN package BH provides a (large) subset of Boost. This package > tricks > | R into believing BH is installed when we just depend on the distribution's > | > > In principle, yes. In practice, not sure. I may miss when Boost updates
Usually this does not happen in a short time frame. Even libboost1.63 is hanging around. We could even switch to "default" boost later. I think the problem is that r-cran-bh has a *higher* version number than what is pulling in. So the dependency relation specified in DESCRIPTION seems to be fullfilled - but it is not. > -- ie > yesterday when I was in a rush I convinced myself that Boost was at 1.62 > (even though I subconsciously knew that wasn't right). You was right - default is at 1.62 and this is actually the problem. > So hard-wiring a version number that we cannot control for is not a good idea. I think if we are aware of this issue it will be visible quite soon and bumping r-cran-bh version is under our control. > | Thanks for maintaining r-cran-bh > > BH has been really useful for R and CRAN. It still creates a tough nut for > us here. I don't see it a tough nut - it works nicely so far and with the proper versioned depends I think we can even solve the said problem. Kind regards Andreas. -- http://fam-tille.de