Re: compiling error

["Jeff C," <nolongersafeto@userealemailsniff.com>]

"Ludovic Brenta" <ludovic.brenta@insalien.org> wrote in message
news:m3brt9b2vn.fsf@insalien.org...
>
> Andrew, I'd suggest you use the GNAT Programming System in addition to
> GNAT.  The GPS is an integrated development environment, in which you
> can edit your program and compile.  It will then let you navigate
> directly to where the error is.
>
> You can find the GPS on http://libre.act-europe.fr/GPS.  If you use
> Debian, you can get the relevant packages from the Ada-France
> repository here:
>


While I think GPS is promising, I have a few concerns about recommending the
academic version for general use (at this time).

There are several "problems" that will almost immediately pop up that are
not addressed on the top level web page. I think these problems are bad
enough to make someone regard this tool in a negative light.

The biggest problem I ran into is that there are features of GPS that at
least appear to assume a gnat 5.0 series
compiler (or at least I have attributed the problems to this). The biggest
being that GPS appears to create project
files that are not valid project files for the 3.15 series compilers. This
ends up making the tool look broken
almost immediately.

The next problem (assuming one works around the one above by building a 5.0
series compiler from CVS) is that
the pretty printer at least appears to be poorly integrated into the tool.
The editor does some level of indent/pretty
printing as you go but if you have existing code you want to reformat you
select pretty print it creates
a new window that is not the original source window to reformat the code.
There are probably reasons why this is
good.. But it feels yucky and requires copy/paste back to original buffer.

The next problem is that the simulated MDI has too many ways to end up with
a window on "top" but no window selected so that when you select compile, it
can not do it because there is no "top" level window. This feels wrong. It
should probably try to prefer the selected window but if no window is
"selected" it should just assume that the window
in the front of the MDI is the one I want to compile..I am sure there are
things that make doing this sound easier than it is
but it is another feature that feels broken from the user point of view.

Unfortunately, under Linux it is probably the best option we have right now.
Under windows, especially for newbie's working on small 2-3 file projects, I
would still stick with AdaGIDE.