[Ada] implement many new Ada 2005 features

[Ada] implement many new Ada 2005 features

[Hyman Rosen]

For those c.l.a. readers who don'tregularly follow the gcc-patches
mailing list of gnu.gcc.org, check the following message:
<http://gcc.gnu.org/ml/gcc-patches/2005-02/msg00420.html>

Good news for Ada devotees, and it should give pause to one or two
overly ambitious n.g. members who are bent on writing their own Ada
compilers from scratch :-)

Re: [Ada] implement many new Ada 2005 features

[Marin David Condic]

Hyman Rosen wrote:

> For those c.l.a. readers who don'tregularly follow the gcc-patches
> mailing list of gnu.gcc.org, check the following message:
> <http://gcc.gnu.org/ml/gcc-patches/2005-02/msg00420.html>
> 
Good to see that there is an Ada 2005 implementation well under way and 
making good progress.


> Good news for Ada devotees, and it should give pause to one or two
> overly ambitious n.g. members who are bent on writing their own Ada
> compilers from scratch :-)

Depends on the reason for wanting to write your own compiler. If you 
want to learn about compiler theory or perhaps try out some techniques 
not done in other implementations or target some machine nobody else is 
interested in supporting, then you might have a really good reason. If 
the objective is to supplant the "industry leader" (is that fair to say 
of Gnat?) then one needs to be really well financed and have some 
anticipated features/benefits that will make the end product better.

Without a "better idea" and lots of resources, I wouldn't try to 
overtake Gnat, et alia. But hobbies, education, theory testing, etc. are 
all good enough reasons to go for it.

MDC

-- 
======================================================================
Marin David Condic
I work for: http://www.belcan.com/
My project is: http://www.jsf.mil/NSFrames.htm

Send Replies To: m   o   d   c @ a   m   o   g
                    c   n   i       c   .   r

     "'Shut up,' he explained."

         --  Ring Lardner
======================================================================

Re: [Ada] implement many new Ada 2005 features

[Hyman Rosen]

Marin David Condic wrote:
> Depends on the reason for wanting to write your own compiler.

Not really. After man-decades of effort, there are still little
corner cases in GNAT that need fixing up. Someone starting over
from scratch is going to lose the benefit of all that experience
and will have to rediscover all those little problems himself.

That's why the temptation to rewrite a system is so pernicious.
A lot of the knowledge lives in the code, and isn't documented
anywhere else. Because of that, the rewriters begin with the
thought that the new system will be cleaner and nicer than the
old one. It's only later when the new system doesn't work that
they discover all the things that the old system can do and the
new one can't.

Re: [Ada] implement many new Ada 2005 features

[Marin David Condic]

I agree that with all the work that has gone into Gnat/gcc that one 
would have a hard time catching up. That's why I said if one wished to 
take over the market dominance owned by Gnat, et alia, one would need to 
be really well financed and truly have some better idea.

But would you suggest that college level compiler classes not assign 
their students to go write a compiler? Probably not. Hence there is a 
reason someone might want to do so. "I want to learn about compilers and 
try out some interesting theories I have..." is a good reason to go 
write one. Taking over the market from "Coca Cola" would probably be a 
bad objective because of the cost & difficulty of actually succeeding. 
But then, Ada itself keeps tilting at that windmill so should we really 
be surprised?

MDC


Hyman Rosen wrote:
> 
> Not really. After man-decades of effort, there are still little
> corner cases in GNAT that need fixing up. Someone starting over
> from scratch is going to lose the benefit of all that experience
> and will have to rediscover all those little problems himself.
> 
> That's why the temptation to rewrite a system is so pernicious.
> A lot of the knowledge lives in the code, and isn't documented
> anywhere else. Because of that, the rewriters begin with the
> thought that the new system will be cleaner and nicer than the
> old one. It's only later when the new system doesn't work that
> they discover all the things that the old system can do and the
> new one can't.

-- 
======================================================================
Marin David Condic
I work for: http://www.belcan.com/
My project is: http://www.jsf.mil/NSFrames.htm

Send Replies To: m   o   d   c @ a   m   o   g
                    c   n   i       c   .   r

     "'Shut up,' he explained."

         --  Ring Lardner
======================================================================

Re: [Ada] implement many new Ada 2005 features

[Robert A Duff]

Marin David Condic <nobody@noplace.com> writes:

> But would you suggest that college level compiler classes not assign
> their students to go write a compiler? Probably not.
...

An Ada compiler is probably not the best choice for a college course. ;-)

- Bob

Re: [Ada] implement many new Ada 2005 features

[Marin David Condic]

Yeah, maybe. Unless the course has a title like "Understanding the Ada 
Programming Language From A Compiler Theory Perspective..." I'm aware of 
a number of compiler courses that at one time or another had students 
implementing mini-Ada compilers. (Probably because the syntax is pretty 
clean for parsing purposes.)

But the point wasn't to design a compiler theory class. It was to 
observe that if someone wants to go off and try implementing an Ada 
compiler for purposes of education or to try out some theories or simply 
because they find compiler implementation an interesting hobby - I say 
don't discourage them. It's a learning exercise that will change them - 
if not the world. I qualified the comment that if the goal was to 
supplant something like Gnat, then its tilting at windmills and not 
really likely to succeed. The effort might be better spent working on 
some interesting new direction for Gnat to go in. (Porting to new 
hardware, adding a new capability, etc.)

MDC

Robert A Duff wrote:
> 
> An Ada compiler is probably not the best choice for a college course. ;-)
> 
> - Bob

-- 
======================================================================
Marin David Condic
I work for: http://www.belcan.com/
My project is: http://www.jsf.mil/NSFrames.htm

Send Replies To: m   o   d   c @ a   m   o   g
                    c   n   i       c   .   r

     "'Shut up,' he explained."

         --  Ring Lardner
======================================================================

Re: [Ada] implement many new Ada 2005 features

[Randy Brukardt]

"Hyman Rosen" <hyrosen@mail.com> wrote in message
news:VvZPd.21381$ya6.16171@trndny01...
...
> Good news for Ada devotees, and it should give pause to one or two
> overly ambitious n.g. members who are bent on writing their own Ada
> compilers from scratch :-)

Well, some of us have our own Ada compilers (Janus/Ada in my case), and I
certainly hope that you are not advocating the elimination of all other Ada
technology. I can't think of anything more certain to kill Ada than to have
only one source for compilers - that would mean that one size would have to
fit all, and that is unlikely to be true.

Certainly, Janus/Ada is not going to compete in every environment with GNAT
(it's certainly not suitable for hard real-time work, for instance), but it
keeps alive techniques that have disappeared elsewhere: real generic code
sharing, discontiguous objects (no allocate-the-max bs here), and so on.

I wouldn't suggest that anyone start from scratch, though, because I've
spent about 17 years of my life working on Janus/Ada (along with a number of
other people), and that sort of investment is not something to undertake for
most humans. And a compiler for the "easy" subset of Ada is unlikely to be
interesting.

                            Randy Brukardt

Re: [Ada] implement many new Ada 2005 features

[Ed Falis]

On Thu, 17 Feb 2005 17:00:16 -0600, Randy Brukardt <randy@rrsoftware.com>  
wrote:

> Well, some of us have our own Ada compilers (Janus/Ada in my case), and I
> certainly hope that you are not advocating the elimination of all other  
> Ada
> technology. I can't think of anything more certain to kill Ada than to  
> have
> only one source for compilers - that would mean that one size would have  
> to
> fit all, and that is unlikely to be true.
>
> Certainly, Janus/Ada is not going to compete in every environment with  
> GNAT
> (it's certainly not suitable for hard real-time work, for instance), but  
> it
> keeps alive techniques that have disappeared elsewhere: real generic code
> sharing, discontiguous objects (no allocate-the-max bs here), and so on.
>
> I wouldn't suggest that anyone start from scratch, though, because I've
> spent about 17 years of my life working on Janus/Ada (along with a  
> number of
> other people), and that sort of investment is not something to undertake  
> for
> most humans. And a compiler for the "easy" subset of Ada is unlikely to  
> be
> interesting.
>
>                             Randy Brukardt

Hear! Hear!

- Ed

Re: [Ada] implement many new Ada 2005 features

[Larry Kilgallen]

In article <rYSdnVHyaq8IvIjfRVn-3w@megapath.net>, "Randy Brukardt" <randy@rrsoftware.com> writes:
> "Hyman Rosen" <hyrosen@mail.com> wrote in message
> news:VvZPd.21381$ya6.16171@trndny01...
> ...
>> Good news for Ada devotees, and it should give pause to one or two
>> overly ambitious n.g. members who are bent on writing their own Ada
>> compilers from scratch :-)
> 
> Well, some of us have our own Ada compilers (Janus/Ada in my case), and I
> certainly hope that you are not advocating the elimination of all other Ada
> technology. I can't think of anything more certain to kill Ada than to have
> only one source for compilers - that would mean that one size would have to
> fit all, and that is unlikely to be true.


Consider the case of Mumps/M where the number of vendors was so reduced
by vendors buying each other out that there is no longer an international
standard due to lack of vendor interest.

Re: [Ada] implement many new Ada 2005 features

[Colin Paul Gloster]

On Wed, 16 Feb 2005, Robert A Duff wrote:

"Marin David Condic <nobody@noplace.com> writes:

> But would you suggest that college level compiler classes not assign
> their students to go write a compiler? Probably not.
...

An Ada compiler is probably not the best choice for a college course. ;-)"

For an entirely different reason a compilers lecturer I used to have had a 
problem with Ada: he believed that structural equivalence and compatiblity 
of types (i.e. weaktyping) was superior to strong typing. Yikes!

Re: [Ada] implement many new Ada 2005 features

[Colin Paul Gloster]

On Thu, 17 Feb 2005, Randy Brukardt wrote:

"[..]

Certainly, Janus/Ada is not going to compete in every environment with GNAT
(it's certainly not suitable for hard real-time work, for instance), [..]

[..]"

However you do have the capability to port your cross development tools to 
target an 8bit processor don't you? Some decade I may want to fund a 
compiler for some poorly supported target I might be stuck with, and I 
might get worried if you could not do a realtime executive for me.

Re: [Ada] implement many new Ada 2005 features

[Randy Brukardt]

"Colin Paul Gloster" <Colin_Paul_Gloster@ACM.org> wrote in message
news:Pine.WNT.4.62.0502271630490.-706617@PC3...
> On Thu, 17 Feb 2005, Randy Brukardt wrote:
>
> "[..]
>
> Certainly, Janus/Ada is not going to compete in every environment with
GNAT
> (it's certainly not suitable for hard real-time work, for instance), [..]
>
> [..]"
>
> However you do have the capability to port your cross development tools to
> target an 8bit processor don't you? Some decade I may want to fund a
> compiler for some poorly supported target I might be stuck with, and I
> might get worried if you could not do a realtime executive for me.

Real-time, yes. Hard real-time, I don't think so.  The primary issue is that
hard real-time systems usually need strict control over memory management as
well. Janus/Ada generally uses dynamic allocation (heap, that is) for
anything not statically sized, which (because of code sharing) includes
function calls and pretty much anything in generics. So writing sane Ada
code that completely avoids dynamic allocation is nearly impossible with
Janus/Ada, and that would make it inappropriate for some purposes. (We'll
probably eliminate some of the dynamic cases from function calls when we
implement extended return statements, and our optimizer already eliminates
many more, but I don't think that the general issue will disappear. Of
course, with enough money, anything is possible...)

                         Randy Brukardt.

Re: [Ada] implement many new Ada 2005 features

[tmoran]

> Real-time, yes. Hard real-time, I don't think so.
Of course with today's CPUs some things that used to be difficult to
do in real time are now easy.  (note Java)

Re: [Ada] implement many new Ada 2005 features

[Vinzent 'Gadget' Hoefler]

tmoran@acm.org wrote:

>> Real-time, yes. Hard real-time, I don't think so.
> Of course with today's CPUs some things that used to be difficult to
> do in real time are now easy.  (note Java)

Still doesn't qualify for "hard real-time". The notion of "immortal
memory" and no heap realtime threads" didn't go into the RTSJ just for
fun.


Vinzent.

-- 
worst case: The wrong assumption there actually is one.

Re: [Ada] implement many new Ada 2005 features

[Colin_Paul_Gloster]

References: <1ZqdnRrgeMe3Br7fRVn-qg@megapath.net> <ScSdnSWjepOPLb7fRVn-hQ@comcast.com>

Note that aicas's work on developing realtime Java is being based on
RAVENSCAR.

The heavyweight processors needed for QNX and VxWorks do not seem to
provide hard enough guarantees to Pete Fenelon's satisfaction
( news:news:4arvuc.h91.ln@fenelon.com , "Re: OS9, hard or soft RTOS?",
Wed, 16 Feb 2005 16:10:44 +0000) (same goes for OS9 by the way). Which makes
one wonder why someone in the Canadian Space AGency ranked QNX as the best
RTOS for the Canadian Space Agency's needs after surveying approximately 48
RTOSes (presented at DASIA 2003)?