GNAT for DOS

GNAT for DOS

[Miguel Miranda]

G'day,

Does or will GNAT for DOS support tasking in the future?


Miguel Miranda
STO
UTS
OZ (Australia)

gnat for DOS

[Al Christians]

There is a note in the gnat archive with the DOS version that says
that the gnat make is broken under DOS.  Does this mean that gnat is
not currently useful to develop programs that will run under DOS, or
is there any way around this problem?  Are there any ideas about of
when the expected remedy to this difficulty will be available? 

Al

Re: gnat for DOS

[Michael Feldman]

In article <317A8B31.CFD@easystreet.com>,
Al Christians  <achrist@easystreet.com> wrote:
>There is a note in the gnat archive with the DOS version that says
>that the gnat make is broken under DOS.  Does this mean that gnat is
>not currently useful to develop programs that will run under DOS, or
>is there any way around this problem?  Are there any ideas about of
>when the expected remedy to this difficulty will be available? 

gnatmake is "broken" in the current (3.01) version of GNAT under DOS
because the djgpp version 1 memory extender loads a copy of itself into
conventional RAM for each program spawned, at 130k per copy. This eats
memory pretty fast during the bind/link phases using gnatmake. We
decided not to distribute gnatmake for this reason.

Workaround? Just compile the individual units individually, instead
of using gnatmake. No big deal unless your program has zillions of
units.

That said, I should point out that a 3.03 release, compiled for djgpp
version _2_ (which uses dpmi memory and has no more of the above
problems) is quite near at hand (a week or so). Better still, this
release has a tasking runtime! We've done some pretty extensive
testing and think this will be a very good release. gnatmake works
just peachy keen. So does the tasking stuff.

The DOS build for djgpp v2, and the completion of the tasking runtime, 
was done mostly by Doug Rupp of ACT, to whom all GNAT/DOS fans will owe
thanks. The final packaging and integration with a new AdaCAPS release
is happening at GW. 

Stay tuned; we'll post an anouncement here, of course.

Mike Feldman
------------------------------------------------------------------------
Michael B. Feldman -  chair, SIGAda Education Working Group
Professor, Dept. of Electrical Engineering and Computer Science
The George Washington University -  Washington, DC 20052 USA
202-994-5919 (voice) - 202-994-0227 (fax) 
http://www.seas.gwu.edu/faculty/mfeldman
------------------------------------------------------------------------
       Pork is all that money the government gives the other guys.
------------------------------------------------------------------------
Ada on the WWW: http://lglwww.epfl.ch/Ada/ or http://info.acm.org/sigada/
------------------------------------------------------------------------

Re: gnat for DOS

[Weston T. Pan]

mfeldman@seas.gwu.edu (Michael Feldman) writes:

>In article <317A8B31.CFD@easystreet.com>,
>gnatmake is "broken" in the current (3.01) version of GNAT under DOS

Isn't there a newer version? I could have sworn I downloaded some files 
called gnat-3.03-dos* . However, they came with djgpp 1.12 and don't 
support tasking.

>That said, I should point out that a 3.03 release, compiled for djgpp
>version _2_ (which uses dpmi memory and has no more of the above
>problems) is quite near at hand (a week or so). Better still, this
>release has a tasking runtime! We've done some pretty extensive
>testing and think this will be a very good release. gnatmake works
>just peachy keen. So does the tasking stuff.

Wonderful!

>The DOS build for djgpp v2, and the completion of the tasking runtime, 
>was done mostly by Doug Rupp of ACT, to whom all GNAT/DOS fans will owe

If you see this, Thank you Doug!

>thanks. The final packaging and integration with a new AdaCAPS release

I really like this editor. I hope the new one supports color syntax and 
doesn't have that 64k limit.

>is happening at GW. 

>Stay tuned; we'll post an anouncement here, of course.

Any plan for interrupt support?

Re: gnat for DOS

[Robert Dewar]

iAl asked

"There is a note in the gnat archive with the DOS version that says
that the gnat make is broken under DOS.  Does this mean that gnat is
not currently useful to develop programs that will run under DOS, or
is there any way around this problem?  Are there any ideas about of
when the expected remedy to this difficulty will be available?"

It is perfectly possible to use GNAT without an automatic make
facility, though it is certainly more convenient to have it available.
The reason that gnatmake does not work with the current DOS version
is that there is not enough low memory space with DJGPP version 1.

Ada Core Technologies will shortly be releasing a new DOS version, which
will also be available in Mike Feldman
's EZLoad packaging that fixes this problem (by using DJGPP 2), and also
fully implements tasking.

P.S. Even with the new version, you are well advised to move to a real
32 bit operating system (OS/2, NT. Solaris 86, Linux, FreeBSD, NetBSD,
Nextstep) to do any serious Ada development on a PC. Of these at least
OS/2, Linux and NT support tasking fully.

Re: gnat for DOS

[Scott H. James]

begin Dewar

  P.S. Even with the new version, you are well advised to move to a real
  32 bit operating system (OS/2, NT. Solaris 86, Linux, FreeBSD, NetBSD,
  Nextstep) to do any serious Ada development on a PC. Of these at least
  OS/2, Linux and NT support tasking fully.

end Dewar


Sigh.

While your comment is perhaps relevant to GUI and OS-application
programming, with regards to embedded programming your repeated smirks at DOS
smack of either gross ignorance or wishful thinking.  Due to DOS's relatively
low system overhead, straightforward access to hardware, and the
proliferation of ROM DOS chips, rest assured DOS is and will remain with us
as a low-end embedded OS option for many years to come.  To continually bash
DOS does nothing for Ada and perhaps, in the future, should be relegated to
comp.os.advocacy or alt.windows.and.mice.are.really.cool or some
other such group.

Thanks,
Dr. Scott

Re: gnat for DOS

[Robert Dewar]

Scott James said:

"While your comment is perhaps relevant to GUI and OS-application
programming, with regards to embedded programming your repeated smirks at DOS
smack of either gross ignorance or wishful thinking.  Due to DOS's relatively
low system overhead, straightforward access to hardware, and the
proliferation of ROM DOS chips, rest assured DOS is and will remain with us
as a low-end embedded OS option for many years to come.  To continually bash
DOS does nothing for Ada and perhaps, in the future, should be relegated to
comp.os.advocacy or alt.windows.and.mice.are.really.cool or some
other such group."

You miss my point, I am not talking about DOS as a target environment. It
is quite true that since DOS is essentially completely transparent at run
time that if what you really want to do is to run on the bare board without
any OS intervention, DOS is perfectly reasonable.

My comment applies to DOS as a *development* environment, where it is indeed
lacking, since running on a DOS extender under DOS is always problematic
compared to using a proper 32-bit operating system.

Note that GNAT in any case does not run under DOS directly, nor do programs
that it generates. Instead they run in protected mode on a 32-bit DOS
extender. Your comments about embedded programming thus don't apply
directly, since in any case you do NOT have straightforward access to
the hardware, and here is no question of running in the context of
ROM DOS chips. 

It remains good advice for people doing serious development using GNAT
on a PC to choose one of the many 32-bit operating systems that is
supported. It is certainly true that DJGPP 2.0 alleviates some of the
worst characterstics of the DOS environment (notably the exhuastion of
low memory that made the use of gnatmake impossible), but still there
are limitations.

So my recommendations here have nothing whatever to do with operating
system advocacy, and are quite relevant to this group. Although it is
certainly possible to run GNAT under DOS, and even to build large
programs in this environment, you are buying yourself trouble. THis
is more a comment on GNAT and its system requirements than on DOS
per se. FOr example, the 8+3 character limits on file names are merely
a minor nuisance for most uses of DOS, but for writing large programs
using GNAT, this file naming restriction can be very bothersome.

That being said, we are continuing to enhance the DOS version. The next
release incorporates tasking, and also uses DJGPP 2.0 (so gnatmake should
finally be available).

Re: gnat for DOS

[Tom Griest]

sjames@wam.umd.edu (Scott H. James) writes:
>
>begin Dewar
>
>  P.S. Even with the new version, you are well advised to move to a real
>  32 bit operating system (OS/2, NT. Solaris 86, Linux, FreeBSD, NetBSD,
>  Nextstep) to do any serious Ada development on a PC. Of these at least
>  OS/2, Linux and NT support tasking fully.
>
>end Dewar
>
>
>Sigh.
>
>While your comment is perhaps relevant to GUI and OS-application
>programming, with regards to embedded programming your repeated smirks at DOS
>smack of either gross ignorance or wishful thinking.  Due to DOS's relatively
>low system overhead, straightforward access to hardware, and the
>proliferation of ROM DOS chips, rest assured DOS is and will remain with us
>as a low-end embedded OS option for many years to come.  To continually bash
>DOS does nothing for Ada and perhaps, in the future, should be relegated to
>comp.os.advocacy or alt.windows.and.mice.are.really.cool or some
>other such group.

I don't think Robert was "bashing DOS".  His comment was advice based
on the fact that you have to be running 32-bits to run GNAT and that it
makes sense to use a "native" 32-bit OS instead of a 16-bit OS with
a 32-bit extender on it.  NOTE: the word "development" in his quote.
As for target systems....  knock yourself out.  We are developing
on WinNT and targeting a 2"x3" 386EX board (and have targeted PC's
running DOS too).  Of course if you only have one box, and don't
want to reboot often, then DOS host/target is probably the way to go.
You just have to judge for yourself if your time lost by using DOS
as a development system is worth less than then the cost of a
second "embedded DOS box" (probably under $500).

-Tom

GNAT for DOS

[fabien_bousquet]

I am using GNAT for DOS 3.10p.
My program is not compiling with this version because of a GNAT bug.
It is compiling on NT with GNAT 3.13p.
But I need to compile it on an old PC with DOS.
Do you know where I could find a more recent release of GNAT for DOS or
some patches ?

Thanks,
        Fabien Bousquet


Sent via Deja.com http://www.deja.com/
Before you buy.

Re: GNAT for DOS

[Gautier]

> Do you know where I could find a more recent release of GNAT for DOS or
> some patches ?

There is GNAT 3.12p on that page:

  http://members.nbci.com/gnatlist/

HTH
______________________________________________________
Gautier  --  http://members.nbci.com/gdemont/gsoft.htm

Re: GNAT for DOS

[fabien_bousquet]

In article <39D9C117.88B0886@maths.unine.ch>,
  Gautier <gautier.demontmollin@maths.unine.ch> wrote:
> > Do you know where I could find a more recent release of GNAT for
DOS or
> > some patches ?
>
> There is GNAT 3.12p on that page:
>
>   http://members.nbci.com/gnatlist/
>
> HTH
> ______________________________________________________
> Gautier  --  http://members.nbci.com/gdemont/gsoft.htm
>
Thanks !!!



Sent via Deja.com http://www.deja.com/
Before you buy.

Re: GNAT for DOS

[Gautier]

> Thanks !!!

You are welcome. BTW the magician who builds the DOS port
will produce a 3.13 one of these nights... BTW 2: is a
"GNAT 3.135" bugfix version in the works ? There seems
to be fresh bugs - e.g. preventing CLAW to compile.

G.

Re: GNAT for DOS

[Robert Dewar]

In article <39DB7777.472843BE@maths.unine.ch>,
  Gautier <gautier.demontmollin@maths.unine.ch> wrote:
> > Thanks !!!
>
> You are welcome. BTW the magician who builds the DOS port
> will produce a 3.13 one of these nights... BTW 2: is a
> "GNAT 3.135" bugfix version in the works ? There seems
> to be fresh bugs - e.g. preventing CLAW to compile.

We are beta testing prereleases of 3.14a at this stage with
several customers. Eventually there will be a 3.14a release
and following that a 3.14p release. We never give advance
schedules for public releases of GNAT.

Robert Dewar
Ada Core Technologies


Sent via Deja.com http://www.deja.com/
Before you buy.

Re: GNAT for DOS

[Robert Dewar]

In article <39DB7777.472843BE@maths.unine.ch>,
  Gautier <gautier.demontmollin@maths.unine.ch> wrote:
> > Thanks !!!
>
> You are welcome. BTW the magician who builds the DOS port
> will produce a 3.13 one of these nights... BTW 2: is a
> "GNAT 3.135" bugfix version in the works ? There seems
> to be fresh bugs - e.g. preventing CLAW to compile.


One thing that would be helpful is to clarify how complete
the DOS port is, in particular does it implement full tasking
at this stage?


Sent via Deja.com http://www.deja.com/
Before you buy.

GNAT for DOS

[chris.danx]

I've downloaded GNAT for DOS but i'm not sure which version of djgpp i need to
run it.  I've got 3.13alpha and i've checked the current verson of djgpp --
2.03.  Has anyone used it and if so which version of DJGPP do i need?  And where
do i get it?


Thanks,
Chris

p.s. i looked in the windows api ref and win9x doesn't support deviceio thingy.
{ Probably due to it's naff security mechanisms -- which are nothing like
t's!  }  Hence the need for the dos compiler!