Re: Building Matreshka on Windows

[Simon Wright <simon@pushface.org>]

"Dmitry A. Kazakov" <mailbox@dmitry-kazakov.de> writes:

> On 04/11/2016 09:57, G.B. wrote:
> [...]
>> Better suited to the task of configuring software translation will be
>> a typed, declarative language for handling rules and constants, I'd
>> think. A language that "knows" about multiple languages like gnatmake
>> knows about Ada. So, it wouldn't be either Ada or Python.
>
> And this is the key point. Should it be a language? Really? If AdaCore
> knows these object languages (Ada, C, C++, Brainf*ck etc), why do they
> need a language to express that knowledge? Why do they need to
> communicate that knowledge to the poor end user? As a programmer I
> have no interest whatsoever in this knowledge. Why don't they let me
> alone?

I just don't understand what makes you say this. If you are happy with
gnatmake without GPRs, carry on and use it, with all the mess of -A*
swithces that you will certainly need a makefile or other script to
remember for you. gprbuild will work just as well, except of course that
you do have to have a project file. Why on earth would you need to look
at $prefix/share/gprconfig?

> Note this same question arises with handling Ada projects. You could
> try to use a language like GNU make to maintain an Ada project. Or you
> could have a non-language tool like gnatmake or AdaGIDE to handle
> that. What works better?

gprbuild, of course.

> You might say they wanted it be open-ended so that I could add the
> Whitespace language to the list. The answer is no. I am not interested
> in doing this and the percentage of people who potentially may differ
> and would willingly become programmers of the configuration language
> does not pay off the miseries the huge majority of "normal" people
> have.

I think the people who have problems such as you hint at with gprbuild
are not "normal".

Of course, one might think that people who want to program in Ada aren't
normal, either ...