Re: wlroots backport?

Hi,

* Michael Tokarev <[email protected]> [230818 14:04]:
> 18.08.2023 02:47, Nick Hastings пишет:
> > Hi Michael,
> > 
> > * Michael Tokarev <[email protected]> [230815 02:55]:
> > > 
> > > There's an issue with wlroots and backporting: it is the soname of libwlroots
> > > library.
> > 
> > can you please explain or point me to an existing explanation of this
> > problem?
> > 
> > I'm curious because I have backported wlroots 0.16.2 from testing/unstable
> > to bookworm. The produced libwlroots-dev and libwlroots11 packages seem
> > to happily coexist with the libwlroots10 that ships with bookworm.
> 
> There you go, this is where the problem is.  By just back-porting wlroots,
> you gain nothing for bookworm, the back-ported lib isn't being used by
> anything, until...
> 
> > > We don't gain much by backporting just the wlroots, the users of
> > > the library has to be rebuild or backported too.
> > 
> > With the backported wlroots I'm able to backport sway from
> > testing/unstable and also able to build/package/use the latest version
> > of my wayland compositor of choice[1,2]. These backports required no
> > modifications to the source packages.
> 
> ..until something *else* is backported too which uses the backported
> wlroots, like sway.  So basically, if we're about backporting this,
> we should think about backporting, for example, sway, with wlroots
> as a pre-dependency, not wlroots alone.  *That* will make sense.
> Backporting wlroots by itself does not.

Understood.

Interestingly, just over 24 hours ago a sway backport enquiry was
posted:
https://lists.debian.org/debian-backports/2023/08/msg00019.html

I am still curious about the soname issue, which I understood to be
independent of the this.

Cheers,

Nick.