Re: AdaCore's survey regarding the future of GNAT Community Edition

[Simon Wright <simon@pushface.org>]

Andreas ZEURCHER <ZUERCHER_Andreas@outlook.com> writes:

> On Wednesday, August 19, 2020 at 9:29:05 AM UTC-5, Simon Wright wrote:
>> Roger Mc <roge...@gmail.com> writes: 
>> 
>> > I did try to build gps from the current Adacore community source but 
>> > was unsuccessful. The main problem being that Adacore seem to be in 
>> > the midst of doing the necessary upgrade from Python 2 to Python3. I 
>> > did attempt to do Python3 modifications myself but eventually got to a 
>> > stage where I could proceed no further
>> I've reached the same stage. I can manage some of the 2-to-3 fixes (not 
>> the one in gobject-introspection, though), but the real problem for me 
>> is that there isn't a consistent complete set of sources, and some 
>> aren't provided on the Adacore community site (e.g. pygobject, langkit, 
>> libadalang, libadalang-tools, ada_language_server). And, so far as I can 
>> see, langkit (20.2) isn't consistent with libadalang (20.2). And, my 
>> Python venv has got screwed. 
>> 
>> Netflix & Twitter.
>
> If multiple well-skilled people cannot build a GPL-licensed source
> code with the source code as provided and instructions as provided,
> wouldn't that be a black-&-white flagrant violation of the GPL?  The
> natural conclusion seems to be: either the source code provided
> mismatched or the narrative instructions to build were omitting some
> secret-sauce, either of which was an unintentional or intentional
> preventative of success.  The unintentionality versus intentionality
> would be able to be determined only after the fact by observing the
> root-cause of the preventative of successful building once that root
> cause is discovered/reported.  This irreproducibility is both notable
> and highly interesting.

Well - I have every sympathy with people who make a binary release and
then move on in staggered stages, aiming for another binary release in a
year's time. The sort of problem you encounter is that the the version
of gobject-introspection on the CE site won't compile with Python 3.8.5,
because of the removal of the DL_EXPORT macro that was deprecated with
Python 2.3; while libadalang _requires_ Python 3.8.5. The latest glib
uses Yet Another Build Tool (meson/ninja), and the script doesn't export
a header required by gtk-3.14+ ... it's not so much DLL Hell as a
version compatibilty tightrope.