[postgis-devel] liblwgeom versioning

[Sandro Santilli]

On Fri, Oct 09, 2015 at 10:36:24AM -0400, Greg Troxel wrote:
> 
> Sandro Santilli <strk at keybit.net> writes:
> 
> > On Thu, Oct 08, 2015 at 09:23:28AM -0400, Greg Troxel wrote:
> >
> > I did, and it would suggest me to use 5:0:0 (decrement of CURRENT is
> > never documented as a possibility). It's just a silly thing that it
> > annoys me to see a jump from "liblwgeom-2.2.so.2" to "liblwgeom-2.2.so.5"
> > ...  
> 
> That seems goofy but I can't argue that you are wrong...   It seems
> obvious that .so.2 with an ABI change should change to .so.3.

We might not be seeing the other side of the moon: some other system
where 4:2:0 does not map to .so.2, for example (I think I saw
something like that on a FreeBSD in the past but don't have access
to it anymore).

> I only sort of follow you, but I get it that combining things that
> should be separate get to be trouble.
> 
> Perhaps it's then a good plan of leaving lwgeom-X.Y.so for now, and
> dropping the -X.Y when it becomes a first-class separate package, with
> its own tarball, release number, pkgconfig, etc.  Which is the path you
> are heading down, I think, eventually.

Yes, agreed.

> >> >   ./configure --without-pgconfig
> >> 
> >> So after doing that, then 'make' and 'make install' will only build
> >> lwgeom?
> >
> > Yes.
> 
> Sounds good; then I could make a liblwgeom package easily.  But until
> some other package needs it, I probably won't.

I believe next spatialite release will be using it.
And maybe you'd be interested in packaging this too:
https://github.com/strk/fineltra/

There's also a QGIS plugin, I've understood, but don't know the details.

> (I am building 2.2 on NetBSD, and running into packaging nits about perl
> - will send a regression test report later.)

Thanks. Let's make 2.2.1 be delicious !

--strk;