On Sun, Jul 14, 2024 at 08:19:29PM +1000, Ben Finney wrote: > On 14-Jul-2024, Paul Gevers wrote: > > > You should only upload the source, not the arch:all binary. > > Thank you, I'll do that. But I'm still concerned about a couple of issues > this raises: > > That admonition appears to conflict with a different policy I'd learned: > > The package upload should include a built binary package, to demonstrate > that it *can* build and to lower the incidence of uploads that are destined > to fail to build from source. > > Has that policy been rescinded?
I am not aware that this is written down anywhere. > If not, how is it compatible with the > advice you cited? TTBOMK the rules currently are: source-only uploads are perfectly fine and desirable. In a few exceptional cases, binaries must be provided; these are (at least): introduction of new source package (= "NEW"), introduction of new binary package (= "binNEW"), upload of a non-free package that is not auto-built. > > > Where can I read more to understand what went wrong here and > > > how to not have a package blocked this way? > > > > https://lists.debian.org/debian-devel-announce/2019/07/msg00002.html > > second section. > > Oh, I had interpreted that differently: that the binary packages would not > migrate, but that the source package would be re-built and *that* would > migrate to testing. > > You mentioned that the build daemons do not build “Architecture: all” > packages? Is that a temporary fault, or is it intentional? The buildds can build "Architecture: all" packages. However, large parts of our infrastructure and customs are not prepared to deal with rebuilds of "Architecture: all" packages, so this is not done. I'll point out that any rebuilds on buildds also change the version number, e.g. by appending +b1 (or higher). Hope this helps, Chris PS: thinking about all this a bit more, ISTM we are missing the "throw away" parts.
