Am Wed, Oct 02, 2024 at 08:34:24PM +0900 schrieb Charles Plessy: > Le Wed, Oct 02, 2024 at 12:53:56PM +0200, Michael R. Crusoe a écrit : > > > > So I propose to add a "architecture-is-little-endian" build-dep to the > > following packages and request the removal of their s390x builds from the > > archive:
+1 > I am all for it but I would like to propose something bigger. > > Let r-bioc-biocgenerics (or r-bioc-biocbase) provide a virtual package > called r-bioc-supported-architectures available for amd64 and arm64 > only, and make all r-boc-* depend on it at the next BioC release (which > takes place this month). It might possibly a good idea to fix the architure depencency inside r-bioc-biocgenerics (or r-bioc-biocbase). However, there are new interesting architectures coming up (like riscv64) and excluding those without good reason is IMHO not rectified by the problems we've seen in the past. > People who want r-bioc-* packages on other architectures can first work > on portability issues outside the Debian main archive, and can create > a local version of r-bioc-supported-architectures to enable building > BiocC on what they want. If the mass-build looks sustainable we can > consider add, maybe after consulting BioC upstream about future > prospects. > > I think that bioinformaticians can aim for other architectures if they > have appetite for the work it represents, and portrers can aim for > bioinformatics if they have the workforce for it, but by default we > should not build these packages everywhere because it puts the whole > burden on mostly just us. I think architecture-is-little-endian on r-bioc-biocgenerics will solve our current problems without creating extra work for future architectures. I'm also scared by the extra work we have to remove all those architectures that are currently building and testing fine. Kind regards Andreas. -- https://fam-tille.de
