Ludovic Courtès writes: Hi Ludo!
> Olivier Dion <[email protected]> skribis: > >> On Tue, 10 Mar 2026, Ludovic Courtès <[email protected]> wrote: >>> Put differently, when would one ask BLUE to create a tarball instead of >>> using ‘git archive’? >> >> I see some cases: >> >> - The project is not using Git but another VCS. >> >> - Git archive is not flexible enough for what is required >> >> - The wanted compression format is not supported by Git >> >> Otherwise, `git-archive(1)' is flexible enought to do a lots of things. > > Yeah and it emits plain tar, so one can use whatever compressor they > want. > > My question in the context of this thread is whether BLUE might end up > recreating ‘make dist’, which is convenient in many ways but leads > people to distribute tarballs that contain pre-built artifacts. For some part this is my doing. Sergio wrote the initial tarball implementation, and I argued that if "we" want to convince (autotools) projects to switch over to BLUE, it helps if we offer them to do so by offering similar features. I did add patches to make source tarball creation reproducible by default, which is already a deviation from autotools. But most of this is "method". And meanwhile, you convinced me to not even generate ChangeLog files for tarballs. > I think we should try to discourage tarballs that contain pre-built > artifacts. In practice, if one is expected to have BLUE installed to > build a given package, a tarball doesn’t bring much anyway compared to a > VCS checkout. This isn't about "method", it's policy. And although I agree it would be good for BLUE to advocate this (and other sane) policy, it might be good if the user has the mechanism to do as they see fit? Greetings, Janneke -- Janneke Nieuwenhuizen <[email protected]> | GNU LilyPond https://LilyPond.org Freelance IT https://www.JoyOfSource.com | Avatar® https://AvatarAcademy.com
