[Development] Future of QBS

Kevin Kofler kevin.kofler at chello.at
Mon Oct 16 04:42:48 CEST 2017

Christian Gagneraud wrote:
> I would resume this post as "I love CMake, CMake is the only way.
> You're all wrong."
> This post doesn't explain anything, doesn't gives any analysis, no
> comparison, no argument whatsoever, nothing.

It makes one important point (and elaborates it to great lengths): developer 
familiarity. Even if QBS were actually a lot better than CMake (something I 
am also very sceptical about), it would still be universally hated simply 
because it is not what developers (and distribution packagers!) know.

As a distribution packager, I am really fed up of some upstream projects 
reinventing their own custom build systems (qmake, gyp, gn, qbs, etc.) that 
don't work with our existing packaging boilerplate.

> How many people had the same reaction when clang started?
> Nowadays, clang is actually far superior to gcc, it brought tooling
> like we would never have dared to dream of .

Yet, Fedora packages are still built using GCC and there are no plans to 
change that any time soon. The generated code is simply more efficient.

> Same goes with SVN vs git, now (almost) everyone have given up with SVN.
> SVN was "CVS in better", git is a completely different approach to
> SCM, SVN is now a zombie.

Yet, the git way to do things is not necessarily better. Revision IDs are 
not comparable without having the absolute history. Developers can commit 
their work locally without pushing it, encouraging intransparent 
development. And the learning curve is a lot steeper if you are not used to 
it yet.

That said, git nowadays has the exact same argument going for it as CMake: 
it is what everyone is now used to.

        Kevin Kofler

More information about the Development mailing list