Re: Project structure with Gnat

[Robert Dewar <robert_dewar@my-deja.com>]

In article <rqh95tufse15@corp.supernews.com>,
  "Vladimir Olensky" <vladimir_olensky@yahoo.com> wrote:
> Just create manually all needed dummy .ALI files and that's OK
> Supposed to be very convenient for user I understand.

Well as noted, the way you do this is simply to do the
compilations from that directory. Note that the object/ali
directory is the mapping of a compilation environment as
defined in the RM, so generally the notion is to have the
current directory positioned to the current compilation
environment.

> So it is not convenient to do that from AdaGide. As AdaGide
> comes with GNAT it would be much better to use GNAT/GCC
> switches to redirect output where needed.

AdaGide really is not designed to deal with complex
organizations. Remember it was intended for simple student
use, and that is the reason we include it with the NT version
of GNAT. Some people are using it for serious development
purposes, and over time, it is being adapted to be better
capable of handling this kind of development, but probably
the best thing is to address specific suggestions to the
author of AdaGide.

> Unfortunately in this area combination of GNAT with GCC do
> things upside down.

Well these sort of features are designed to be convenient
for the way our customers use GNAT, and so far this has not
particularly been an issue, of course, as above, our customers
are not generally using AdaGide.

> It is not good when  system does not help to use itself but
> rather provoke for struggle (using some workarounds).

Well we certainly don't like to see people struggling, and when
our customers find they are struggling over something, we
certainly look at it carefully, but as you very well know
you certainly cannot satisfy all the miscellaneous wishes
on CLA, since many of them are contradictory :-)

The -i switch to gnatmake was incidentally directly from the
design wishes of several major customers, and does EXACTLY
what they want.

> Here again we see the difference between highly successful
> commercial systems (e.g. Delphi, MSVS)

If Delphi is commercially successful, that sure is news to
Borland (of course that is not their name any more, and that's
part of the story).

In any case, I think highly successful commercial systems get
that way from listening to the needs of commercial customers,
and we certainly follow that approach!

> Anything that makes life of user more easy helps to succeed.

Within limits, extra options always make the user's life harder,
because they make the system more complex. You have to make sure
that the gain is worth the additional complexity.

> Anything that annoys user does not help to succeed.
> People in general are very sensitive to small annoying
> inconveniences.

Yes, but people have very different ideas about what is a small
annoying inconvenience. The above statement is absolutely true.
BUT it does not mean that because YOU find something to be a
small annoying inconvenience that it is a weakness in the system
generally. What you may consider an annoying inconvenience may
be seen as unimportant or even desirable by others.

In this particular case, you would like an enhancement so that
-o can be used to redirect the object and ali files (it is a bit
odd to redirect two files in this way, but presumably you can
figure out something). The issue is simple, this just is not a
feature that anyone has asked for. If a customer really needs
this and there is no better way of doing things it would
probably get done!

> Regarding this particular issue it is so trivial thing for
> compiler system developers to provide output  files
> redirection according to their types and
> this is such common need that I do not understand why it is
> not done so far
> by GNAT/GCC.

OK, so you don't see why this is trickier than you think ...
I guess you need to think some more about it :-)


Sent via Deja.com http://www.deja.com/
Share what you know. Learn what you don't.