Re: .version and .tarball-version, git-version-gen

[Bernhard Voelker]

On 1/25/25 13:57, Bruno Haible via Gnulib discussion list wrote:
>    1) Most distributions do that. For example, coreutils in Debian has 3 
patches
>       [1].
>
>    2) Making modifications is the essence of Free Software.
> [...]
> [1] https://sources.debian.org/patches/coreutils/9.1-1/

Of course: any changes and patches are okay ... and our daily work.
As long as they are transparent to the recipients.
It's about reproducibility, integrity and trust.

The distributions keep their downstream patches and build recipes in
version-control, so it's always clear which combination of sources
and patches lead to which binary package version.

If I pass a package to you with the name 'coreutils-X.Y-dirty.tar.xz',
then what would that tell you?  You cannot know and reproduce what it
consists of.

But if I commit my local change, and pass to you that patch and a
package coreutils-9.6-8-gfbfd886e5, then one has a chance to know.
Hence, anonymous, local changes which can change any second and which
are not reproducible are "dirty".

Isn't it like a grocer weighing out 2 kilogram of wheat for you; you see it.
Then he goes into the backward space and comes back with it saying
"I modified it a bit".  He will need to weigh it out for you again before
you can know and trust that it's still 2kg.
In our case, we need a commit.

Anyway, the term "-dirty" comes from an underlying tool we're using,
so I'm wondering if discussing about it in gnulib is the right place.
We're just the messenger.

Have a nice day,
Berny