On Wed, Apr 1, 2020 at 1:10 PM Ivan Zhakov <i...@visualsvn.com> wrote:
> On Tue, 31 Mar 2020 at 23:13, Justin Erenkrantz <jus...@erenkrantz.com> > wrote: > > > > On Tue, Mar 31, 2020 at 2:33 PM Branko Čibej <br...@apache.org> wrote: > > > > > Thanks, Justin. I don't know if we need another 1.3.x release, given > > > that 1.4 will be API compatible and has more robust OpenSSL 1.1.x and > > > LibreSSL support. > > > > > > > My preference would be to just do a small 1.3.x release out with critical > > build/test fixes; but, if I'm alone and folks want to jump to a 1.4.0 and > > the downstream packagers are fine with upgrading to that, I can be > > convinced to just skip and go to 1.4.0; but, there's probably a longer > > turnaround for getting 1.4.0 out the door though. > > > +1. Given the upstream OpenSSL reversion, I think I'm now leaning towards just focusing on a 1.4.0 release. I went through the motions of trying out the CMake process on Windows - I got lost a bit in the dependency chain, but it does seem promising. (Expat, APR, APR-util are all on CMake now; OpenSSL isn't quite yet, AFAICT.) On a related note, I'm intrigued in expressing the dependency chain via Buildstream - which is going to be moving over to Apache (via Petri): https://mail.gnome.org/archives/buildstream-list/2020-April/msg00010.html At a high-level, the new CMake build system would stay the same, but there'd be a set of files to help bootstrap the build process for beginners and downstream package maintainers. I could see that benefiting APR, APR-util, httpd, and Subversion too...so, I think I'm going to take a pass at trying to see what that'd look like. That would make me a bit more comfortable about the switch to CMake for 1.4.0. Here's an example of a Buildstream repos for Apache Arrow (which is similarly complex): https://gitlab.com/BenjaminSchubert/apache-arrow-build-buildstream-demo-2019 Cheers. -- justin