On Mon, 2025-07-28 at 16:01 +0200, Andrea Pappacoda wrote: > On Mon Jul 28, 2025 at 3:39 PM CEST, Mathias Gibbens wrote: > > Hmm, yes this is less than ideal. There appear to be no actual > > reverse dependencies on golang-github-mitchellh-hashstructure-v2-dev, > > while golang-github-mitchellh-hashstructure-dev has 47. > > Oh, that's higher than I though. Are you counting direct dependencies, > or transitive ones too?
I ran `build-rdeps golang-github-mitchellh-hashstructure-dev` in a sid container, which includes both direct and transitive dependencies. > > I'd suggest modifying this bug into a RM request for > > golang-github-mitchellh-hashstructure-v2-dev; CC'ing Arthur who had > > originally uploaded the v2 package for an ACK from him first. > > When hashstructure-v2 was uploaded, hashstructure was still at v1. So > maybe it made sense to have two packages back then? > > Still, hashstructure declares an unversioned import path, while it is > incompatible with packages depending on v1 (such as a package which > a friend of mine is trying to work on as his first package -- no ITP > yet, I haven't yet introduced him to the concept). Shouldn't then > hashstructure declare a /v2 import path, and hence be renamed too? This > was my reasoning when asking to drop the unversioned (v1) package. Most of the golang packages in Debian have unversioned import paths, which does mean that all dependencies in the archive need to use the same major version of the library. Individual golang libraries vary widely in how well they follow semantic versioning. Given that all packages in the archive currently use v2 from the unversioned package, I think that RM'ing golang-github-mitchellh-hashstructure-v2-dev is the correct path forward. Mathias
signature.asc
Description: This is a digitally signed message part
