Depending on how fancy you want to be with pulling some tricks in the deb,
you could also:

1. Use provides, set up some simlinks and sed action, and file a RM
2. Take over the binary package and build something from the new sources
under the old name

I haven't looked into doing this with a go dev package yet though, so grain
of salt etc - I've spent zero time thinking about this critically

Paul


:wq

On Mon, Sep 1, 2025, 2:22 PM Simon Josefsson <[email protected]> wrote:

> Zygmunt Krynicki <[email protected]> writes:
>
> > With this in place, is there a mechanism where one of the packages
> > could be sunset and replaced by the other? Is that desirable? I think
> > the go module replacement functionality can do it at the go level and
> > something similar is technically doable at package build time.
>
> Not really, I think the best we have is something like this:
>
> 1) Get golang-github-canonical-cpuid into Debian.
>
> 2) Update all reverse build dependencies of golang-github-intel-go-cpuid
> and golang-github-klauspost-cpuid to use golang-github-canonical-cpuid
> instead.
>
> 3) Ask for removal of golang-github-klauspost-cpuid to use
> golang-github-canonical-cpuid.
>
> That may appear simple, but there are traps including:
>
> A) Instead of patching projects directly, we should really ask upstream
> to change and use the new project.  This introduce 1) delays, possibly
> many months, and 2) risk that upstream refuses because of Reasons and
> sometimes they are good and sometimes they are bad.  I've done this a
> couple of times, and sometimes upstream comply but it has happened
> several times that they did something else instead, sometimes reasonable
> and sometimes problematic.
>
> B) There may be subtle API/ABI differences between the projects, are you
> sure your project is a full superset of the earlier ones?
>
> C) For how long will you maintain your version?  If one of these
> packages have been around for 5-10 or more years and worked fine, we as
> external people may rightly wonder if your new shiny project won't just
> end up in the same shape 5-10 years down the line.  So please consider
> if you can't just make one of the existing project work for you instead
> of creating another project.
>
> People have different preferences and thinking on these topics.  I
> prefer to package what upstream ships if it can be made to work, and to
> report wishes on upstream to them and let them decide.  This leads to
> slow results, which sometimes frustrate.
>
> /Simon
>

Reply via email to