Re: Building the 2021 source release of GnatStudio

[Shark8 <onewingedshark@gmail.com>]

On Saturday, July 31, 2021 at 6:29:07 AM UTC-6, Dmitry A. Kazakov wrote:
> > The biggest complaint I had about GNATStudio was its instability. I 
> > think that Adacore has made great progress now. It's now a pleasure to 
> > work with.
> Yes, but each new version of GTK can change that. GTK is unstable on 
> both Windows and Linux, it is just as it is. AdaCore can at best work 
> around GTK bugs.
And this is exactly why I said that one should evaluate the consequences of your dependencies: depending on something unstable can easily introduce that instability into your program.
> 
> Though Python is 100% self-inflicted damage.
Agreed.
While Python can be quick and "easy" for prototyping, there are whole classes of errors that any dynamic-typed language possesses which statically-typed languages do not. The only dynamically-typed programming language that I've come across which [arguably] has both the mechanisms and culture addressing those issues is LISP.
>AdaCore could easily 
> implement some Ada script, again, not to confuse with shell. They did it 
> partially with GPR. The GPR compiler could be extended to support a 
> larger variety of expressions.
GPR is a very saddening example.
It's too "stringly-typed", it doesn't leverage the obvious structural Ada ancestry, and because of this the GPR-build tool is kneecapped.
(As an example, if GPR files were a tightly restricted GENERIC package [that is, a subset of Ada s.t. all GPR files were valid Ada], with build-options as formal-parameters, you could use the GPR-build tool to automatically generate menus and guide the user through a build.)
> 
> Customers wanting Python will use Eclipse instead of GPS anyway, so that 
> is not an argument either.
I see the reasoning there, but what basis do you have for such a strong assertion?