[cfe-dev] is configure+make dead yet?

[James K. Lowden]

On Wed, 20 Jun 2012 17:13:41 -0700
Nick Lewycky <nlewycky at google.com> wrote:

> Are there developers stuck on systems that don't have a recent
> enough cmake in their most recent release, or maybe are using some
> features from configure+make

I used autoconf to build Clang not because I'm "stuck" on a
system without Cmake, but because I have expertise in autoconf and none
with Cmake.  I've never found Cmake compelling enough to justify
learning a new feature-test and dependency syntax.  

I have company.  Users of autoconf outnumber those of Cmake by 10:1 if
not 1000:1.  

Why is that significant?  Everyone acknowledges that documentation is a
challenge for the Clang project.  Clang is a big project dependent on
another big project, each with a host of options.  The last thing
anyone thinking of building it wants to deal with is an alien build
environment.  Whatever defects autoconf+make may have, they have the
advantage of being familiar and predictable, hence tractable.  They
promote self-help, meaning more people can build and use Clang.  

>From the point of view of the project itself, the trade-off is between
(1) maintaining support for autoconf and (2) making Cmake as easy to
use for the autoconf population.  From comments on this thread, it
seems there are things Cmake can't do and others it doesn't do well.
If and when those are rectified, the work of documenting the use of
Cmake in sufficient detail to make up for (so to speak) the lack of
familiarity would still remain.  

So.  If the goal of the Clang project is to have ever more users of
Clang, it's in the project's interest to support autoconf for the
foreseeable future.  Even if that's *not* the goal, supporting autoconf
might continue to be easier than dealing with a lot of Cmake issues and 
questions on cfe-dev.  

--jkl