3.13.0beta1 ?

[Greg Troxel]

>> On Aug 15, 2024, at 2:21 PM, Mike Taves <mwtoews at gmail.com> wrote:
>> 
>> I'd like to bump the minimum CMake version to 3.16 for the GEOS 3.13
>> series. Now is the right time to do this. I don't have a PR ready, but
>> I can later today.
>> 
>> FWIW, CMake 3.16 is the minimum version for PROJ and GEOS.
>> 
>> See also https://github.com/libgeos/geos/pull/936

"Regina Obe" <lr at pcorp.us> writes:

> I assume that GEOS there is meant to be GDAL?
> What does upgrading CMake buy us.
>
> That ticket you reference was fixed a different way.
>
> I'm still hesitant to up the CMake version just for the sake of upping the CMake version especially so late in the cycle of GEOS 3.13 development.
> If we had done this early own, I would not have an issue.
>
> a) GEOS has no dependencies, so is something people can easily compile themselves unless we go around upping version requirements on them
> b) GEOS is a much simpler project than GDAL and PROJ so has fewer needs
> c) Granted I am less concerned about Ubuntu 20.04  and Debian 10  now that Ubuntu 22.04, 24.04, Debian 11 and Debian 12 are out.
>
> But I want to know what goodies we are going to get out of upgrading CMake.

I am a usual objector to increasing requirements quickly; I typically
want any version of anything that was current in the last 2 years to be
ok. (As in, if X was the recommended version on August 16, 2022, even if
it was released in late 2021, then X should be ok.)

on pkgsrc:

  $ cmake --version
  cmake version 3.30.2

I don't think this is "upgrading cmake".  It's "requiring people that
build geos to have a non-ancient cmake".

Looking up cmake 3.16, it was released on November 26, 2019.  So we're
really close to 5 years.

My usual LTS rant: if someone wants to run software from 2019 for that
long-term stability goodness, they can use old geos too.

And to the other comment: C++ requirements are a much bigger deal.  At
this point it's ok, just barely, to demand C++17.  C++11 has been ok for
a few years.  Requiring anything newer than C++17 is unreasonable.

So while I don't get to vote, I think "cmake >= 3.16.0" is a non-event
and just fine (given that we already have cmake instead of autoconf :-).