Re: ✘asciidoc vs asciidoctor, issue #118

[Greg Troxel]

"Gary E. Miller" <gem@rellim.com> writes:

> Works for me.  I do " scons -c && scons && scons dist".
>
> That not work for you?

That works fine.  What I meant is a scheme to e.g. generate man pages
when making a distfile, and then when building from the distfile using
those pre-built man pages if hte tool isn't there.

By accident, I tried to use slightly old gpsd to build in pkgsrc, a
version that is asciidoctor only, without my accomodation to find our
asciidoctor.  man pages were not built and no pre-built ones were
installed (which is ok).

Then, I used the right tarball and got the same set of manpages I used
to get (with asciidoc probably).

>> Overall, unless a tool is painful (because it's in a non-portable
>> language like Haskell or Java, which this isn't, or because it uses
>> QT, which requires many hours of CPU time to build), I lean to just
>> requiring the tool and keeping the world simpler.
>
> I find installing AsciiDoctor on Raspberry Pi3 to be very painful.
> No way I want that to be not possible.

Perfectly OK with me to have the option.  I just forced manbuild=yes in
my build script.

On a RPI3 Model B with a slow spining disk and without ruby, configured
to do builds with one job only to optimize for heat, I built the
asciidoctor package from source (hence forcing the build of ruby 2.6).
It took:

  real    56m51.224s
  user    42m55.768s
  sys     12m3.168s

and to install the binary package an additional

  real    0m21.314s
  user    0m7.192s
  sys     0m3.763s

Basically 4 minutes for libyaml, 50 for ruby and 3 for asciidoctor.
Only 25% of what cmake takes :-(

So I can see why you don't want to build it.

signature.asc
Description: PGP signature