More Ganssle on Ada

More Ganssle on Ada

[mjsilva]

http://www.embedded.com/electronics-blogs/break-points/4404928/Ada-2012-redux?cid=Newsletter+-+Whats+New+on+Embedded.com

Lots of good answers to good questions.  And GNAT is being ported to low-end ARM parts - yipee!

Re: More Ganssle on Ada

[Jeffrey Carter]

On 01/17/2013 01:20 PM, mjsilva@scriptoriumdesigns.com wrote:
> http://www.embedded.com/electronics-blogs/break-points/4404928/Ada-2012-redux?cid=Newsletter+-+Whats+New+on+Embedded.com
>
> Lots of good answers to good questions.  And GNAT is being ported to low-end ARM parts - yipee!

AdaCore's response on when to use C: "Maybe it's a small custom chip for which 
the manufacturer doesn't provide anything other than C."

In such cases, there is at least 1 Ada compiler that uses C as its intermediate 
language and the C compiler as its back end. In fact, I think AdaCore has such a 
compiler, obtained in the merger with SofCheck. This doesn't seem like a reason 
to use C to me.

-- 
Jeff Carter
"C++ is like giving an AK-47 to a monk, shooting him
full of crack and letting him loose in a mall and
expecting him to balance your checking account
'when he has the time.'"
Drew Olbrich
52

Re: More Ganssle on Ada

[Randy Brukardt]

The article also claimed that Ada wouldn't be a good choice for a web 
server. Besides the fact that I wrote my own web server in Ada, I suspect 
all of the people using AWS would disagree. Perhaps the author meant web 
*client* (where the choices are "Javascript" or something that compiles into 
Javascript).

                                             Randy.

Re: More Ganssle on Ada

[J-P. Rosen]

Le 17/01/2013 23:32, Jeffrey Carter a �crit :
> AdaCore's response on when to use C: "Maybe it's a small custom chip
>  for which the manufacturer doesn't provide anything other than C."

and Randy:
> The article also claimed that Ada wouldn't be a good choice for a
> web server. Besides the fact that I wrote my own web server in Ada,
> I suspect all of the people using AWS would disagree. Perhaps the 
> author meant web *client* (where the choices are "Javascript" or 
> something that compiles into Javascript).

To which I would add:
> Where there's no reliability-oriented concern whatsoever, (...) I
> personally would have a preference towards Visual Basic.

I was also confused by these responses, but I think in these cases, the
"I" is the paper's author (Jack Ganssle), not AdaCore's opinion. And he
had to show some balance between the various languages.

-- 
J-P. Rosen
Adalog
2 rue du Docteur Lombard, 92441 Issy-les-Moulineaux CEDEX
Tel: +33 1 45 29 21 52, Fax: +33 1 45 29 25 00
http://www.adalog.fr

Re: More Ganssle on Ada

[Patrick]

I am happy to hear there is an ARM port coming, although who knows if it will still cost 20K...

The answers in this article were pretty wacked and disheartening. I really hate C but that does not mean people are not writing terrific code with it. Slamming it because it's old is just stupid. Ada is about 30 years old now, so people could say the same thing about it. The "recommended" languages mentioned are themselves written in C.

Another thing that has been touched on... is Ada only for life critical applications? Adacore seems to think so. Why NOT build a telephone application in Ada. I am learning Ada because it's a feature rich, sensible language not because I am worried about killing people by accident.

Adacore has to chaise profits, I don't blame them for being slow to support certain hardware targets and such but they should represent the language better. If people believe this crap there will be a whole subset of tinkers and novice programmers that will pass over the language and when their dreams reach fruition their projects will not be using Ada, even though it may have been the best tool for the job

Re: More Ganssle on Ada

[Brian Drummond]

On Fri, 18 Jan 2013 07:08:11 -0800, Patrick wrote:

> I am happy to hear there is an ARM port coming, although who knows if it
> will still cost 20K...

There are other ARM efforts in the open source community too...

> Another thing that has been touched on... is Ada only for life critical
> applications? Adacore seems to think so. Why NOT build a telephone
> application in Ada. I am learning Ada because it's a feature rich,
> sensible language not because I am worried about killing people by
> accident.

Agree entirely. 

Where his point is that there IS a startup cost to using Ada, when there 
is infrastructure to support C/Java/whatever - that is certainly true - 
for now, at least.

And as for "wouldn't build a web server application in it" ... AWS is 
easy, reliable, and it's amusing to see all the random IP addresses out 
there asking AWS for phpMyAdmin/scripts/stuff to exploit...

On the whole the article read to me as if the authors were shy or even 
scared of challenging the concensus position. Perhaps that makes business 
sense for Adacore; strengthen their position in their own niche, rather 
than make a play for more ground. I don't know.

> If people believe this crap there will be a whole
> subset of tinkers and novice programmers that will pass over the
> language, 

... but unfortunately, this is the choir ;-)

- Brian

Re: More Ganssle on Ada

[Dirk Craeynest]

>On Fri, 18 Jan 2013 07:08:11 -0800, Patrick wrote:
>> Another thing that has been touched on... is Ada only for
>> life critical applications?  [...]  Why NOT build a telephone
>> application in Ada.  [...]  If people believe this crap there
>> will be a whole subset of tinkers and novice programmers that
>> will pass over the language,

Brian Drummond  <brian@shapes.demon.co.uk> wrote:
>... but unfortunately, this is the choir ;-)

Hence the need to also preach outside of the choir...

See:
http://www.cs.kuleuven.be/~dirk/ada-belgium/events/13/130203-fosdem.html
https://fosdem.org/2013/schedule/track/ada/

Dirk
Dirk.Craeynest@cs.kuleuven.be (for Ada-Belgium/-Europe/SIGAda/WG9 mail)

Re: More Ganssle on Ada

[Robert A Duff]

Patrick <patrick@spellingbeewinnars.org> writes:

> ...is Ada only for life critical applications? Adacore seems to think
> so.

Why do you say so?  AdaCore builds all sorts of non-safety-critical
stuff in Ada.  Compilers come to mind.

So do some of our customers.

- Bob

Re: More Ganssle on Ada

[Georg Bauhaus]

On 19.01.13 15:19, Robert A Duff wrote:
> Patrick <patrick@spellingbeewinnars.org> writes:
> 
>> ...is Ada only for life critical applications? Adacore seems to think
>> so.
> 
> Why do you say so?  AdaCore builds all sorts of non-safety-critical
> stuff in Ada.  Compilers come to mind.

I guess the point is that the article doesn't say so.

Rather, the interviewees talks about niches for Ada, tools, high
integrity and other thing *not* part of AdaCore's compiler (sadly);
they suggest it is wise to *not* consider Ada where other languages
are chosen. Well, if you want people to listen to what you
have to say about that old DoD government failure, doing the above
might do the trick.

They list some technical reasons for not choosing Ada. These include
presence of libraries, or availability of web servers,
richer than a specialized thingie like AWS, I should think - hey,
when would you tell anyone to stop using J2EE because there
is AWS and be considered sane? Or some "embedded" equivalent of
Java for mobile devices? Or not use Objective-C when there is no
Objective-Ada yet?

If you feed on consultancy, a vague praise of everything, showing
openness to other language choices, surely expresses experience, and the
right attitude, both PR-wise and strategically, since if Ada consultants
want to have a foot in  the door (or want less prejudiced readers,
in Ganssle's case), then shouting and denigration don't usually help.

A win-win situation.

Re: More Ganssle on Ada

[Dmitry A. Kazakov]

On Mon, 21 Jan 2013 13:32:02 +0100, Georg Bauhaus wrote:

> On 19.01.13 15:19, Robert A Duff wrote:
>> Patrick <patrick@spellingbeewinnars.org> writes:
>> 
>>> ...is Ada only for life critical applications? Adacore seems to think
>>> so.
>> 
>> Why do you say so?  AdaCore builds all sorts of non-safety-critical
>> stuff in Ada.  Compilers come to mind.
> 
> I guess the point is that the article doesn't say so.
> 
> Rather, the interviewees talks about niches for Ada, tools, high
> integrity and other thing *not* part of AdaCore's compiler (sadly);
> they suggest it is wise to *not* consider Ada where other languages
> are chosen. Well, if you want people to listen to what you
> have to say about that old DoD government failure, doing the above
> might do the trick.
> 
> They list some technical reasons for not choosing Ada. These include
> presence of libraries, or availability of web servers,
> richer than a specialized thingie like AWS, I should think - hey,
> when would you tell anyone to stop using J2EE because there
> is AWS and be considered sane? Or some "embedded" equivalent of
> Java for mobile devices? Or not use Objective-C when there is no
> Objective-Ada yet?
> 
> If you feed on consultancy, a vague praise of everything, showing
> openness to other language choices, surely expresses experience, and the
> right attitude, both PR-wise and strategically, since if Ada consultants
> want to have a foot in  the door (or want less prejudiced readers,
> in Ganssle's case), then shouting and denigration don't usually help.

And, BTW, even in the niche areas like mission critical, it is to expect
new customers coming to you rather from the general purpose camp, than from
people actually specializing there. Those have their preferences already
and won't switch. So if you target the former, you will get mission
critical contracts as well. See how Java did it.

-- 
Regards,
Dmitry A. Kazakov
http://www.dmitry-kazakov.de

Re: More Ganssle on Ada

[Patrick]

 
> A win-win situation.

I think it's more of a win-lose situation.

This group has been a friendly and VERY helpful place for me, I don't like being the dark cloud complaining about other people but hopefully a little bit of criticism will move Adacore's management in a better direction and help the language. This is the last time I will post negative comments in this thread.

So yes, if someone said Ada was a drop in replacement for bash, they would lose credibility and consulting would not work.

However in the Erlang example I was trying to show how Ada could compete due to it's strengths.

Furthermore I am angry about the comments regarding the ARM port. I am guesstimating the move towards ARM occurred 10+ years ago but since Adacore's customers have been slow to move towards it, Adacore is just now working on the ARM port. What kind of business logic is this? Everyone in business wants to use their existing infrastructure to make as much move as possible, after all why invest in new infrastructure if there is still tons of money to be made with what you have.

Adacore writes compilers and runtimes, writing an ARM port is already a part of it's core business. There are billions of devices that use ARM, why wait for EXISTING customers to move towards ARM, why didn't management start the ARM port 10 years ago so that they could get NEW customers. Why wait until competing languages have solidified their market-share??

Basically everything I don't like about the Ada world:
1)GPL or proprietary stuffed down my throat(outside GCC)
2)An embedded language that can't be used for embedded design(as in baremetal)

have to do with Adacore's management, not the language itself, which is near perfect.

Re: More Ganssle on Ada

[Britt]

On Monday, January 21, 2013 9:34:36 AM UTC-5, Patrick wrote:
> Adacore writes compilers and runtimes, writing an ARM port is already a part
> of it's core business. There are billions of devices that use ARM, why wait for
> EXISTING customers to move towards ARM, why didn't management start the ARM 
> port 10 years ago so that they could get NEW customers. Why wait until 
> competing languages have solidified their market-share?? 

I was also puzzled by the answer regarding the ARM port. AdaCore did have a (customer sponsored) ARM port a few years ago (see http://web.archive.org/web/20101201035019/http://www.adacore.com/home/products/gnatpro/supported_platforms/ via the Wayback machine) but they eventually withdrew it after the sponsor chose not to use it after all (or so I heard). AdaCore currently supports the ARM processor in the Lego Mindstorms robot with GNAT GPL.

Re: More Ganssle on Ada

[Lucretia]

Ok, let's sort this once and for all.

Building an ARM Linux compiler is doable now. The only problem with it is that it doesn't use the unwind-arm.[ch] files in gcc/config/arm but does implement exceptions with setjmp/longjmp.

I ave tried to sort this out, it requires someone with more knowledge of the RTS than me. Basically, what needs to be done to sort this out is to move the current DWARF EH code into a new package, and create an ARM equivalent, this is because the standard unwind and the unwind-arm stuff is very different and the data would be wrong on the Ada side.

I submitted a patch to the GCC ML which does the first part and kind of works, but fails in the end. The response from AdaCore was something along the lines "you need to construct and ARM specific exception object," thanks, very helpful.

I have built the meta-ada layer for Yocto also, but this doesn't build shared libs at this time as the compiler just won't do it, not sure why at this time.

I have built arm-eabi-none-elf compilers as well and used it for bare metal programming.

So, it's doable, but limited.

Luke.

Re: More Ganssle on Ada

[Andrew Haley]

Lucretia <laguest9000@googlemail.com> wrote:
> Ok, let's sort this once and for all.
> 
> Building an ARM Linux compiler is doable now. The only problem with
> it is that it doesn't use the unwind-arm.[ch] files in
> gcc/config/arm but does implement exceptions with setjmp/longjmp.
> 
> I ave tried to sort this out, it requires someone with more
> knowledge of the RTS than me. Basically, what needs to be done to
> sort this out is to move the current DWARF EH code into a new
> package, and create an ARM equivalent, this is because the standard
> unwind and the unwind-arm stuff is very different and the data would
> be wrong on the Ada side.
> 
> I submitted a patch to the GCC ML which does the first part and kind
> of works, but fails in the end. The response from AdaCore was
> something along the lines "you need to construct and ARM specific
> exception object," thanks, very helpful.

I remember offering to help someone do this, but it all went quiet.
It's not terribly difficult: a simple worked example of the changes
needed is in
http://gcc.gnu.org/viewcvs/trunk/libjava/exception.cc?view=markup

... just look for ifdef __ARM_EABI_UNWINDER__

Andrew.

Re: More Ganssle on Ada

[Lucretia]

On Thursday, 24 January 2013 12:18:08 UTC, Andrew Haley  wrote:

> ... just look for ifdef __ARM_EABI_UNWINDER__

It's not that simple. It requires Ada RTS changes too and I just couldn't find the other bits I needed to change.

And yeah, it was probably me you offered.

Luke.

Re: More Ganssle on Ada

[Georg Bauhaus]

On 21.01.13 15:34, Patrick wrote:
> Adacore writes compilers and runtimes, writing an ARM port is already a part of it's core business. There are billions of devices that use ARM, why wait for EXISTING customers to move towards ARM, why didn't management start the ARM port 10 years ago so that they could get NEW customers. Why wait until competing languages have solidified their market-share??

http://www.ubuntu.com/devices/phone/operators-and-oems

Re: More Ganssle on Ada

[sbelmont700]

>"The fundamental computation model of Ada is imperative, and most concepts of imperative programming are available with semantics that are very close to those of these other languages."

I think that right there is the problem: Everyone feels it necessary to compare Ada to C, when really they are so totally different.  I've had the displeasure of maintaining a lot of legacy code, ostensibly written in Ada, where Integer and Float are used for everything, access values are used instead of out parameters, tasks are avoided in favor of the underlying O/S model, System.Address is used instead of generics, and the list goes on.  And then they complain about Ada because they still have to spend just as much time debugging!  A C program written in the Ada syntax is still just a crappy C program.

-sb

Re: More Ganssle on Ada

[Patrick]

>
> Why do you say so?  AdaCore builds all sorts of non-safety-critical
>
> stuff in Ada.  Compilers come to mind.
> So do some of our customers.
>
>
>
> - Bob



I am not trying to make comments about what projects Adacore or their customers are involved in. What I am infuriated about is how Adore is representing the language as a whole.

Let's delve further into the comment about not using Ada for a telephone application. What language comes to mind with such an application. Erlang. Hold this thought.....

CouchDB's primary creator made a courageous decision to rewrite his c++ application in Erlang.  Now imagine that Erickson made a statement earlier that Erlang should not be used for databases applications. Would he still have had the confidence to try?

Now the fear, uncertainty and doubt(FUD) floating around with Erlang right now is that it does not scale as advertised and it's concurrency model is not so great.

I have no idea if this is true but let's just say there was a project manager that did and that he/she went out shopping for another language. They would want a language that was had facilities for building high performance, high reliability and massively concurrent applications, sound like a language we know ?????

Only they read that Adacore says Ada is not a good match for this so they move on to the next language.

So when I started out with Ada about 14 months ago I had many assumptions that have been corrected. I thought that since it's part of GCC that I could use it for embedded design and target all the chips GCC supports. Now I understand how complex a job it is to write a runtime and even though I am disappointed it can't really be used for embedded design(as in baremetal on ARM) I forgave Adacore. They have a business to run and if writing runtimes for lots of targets is not in their best interest, they can't do it.

Later on I discovered that they re-licensed a large number of libraries GMGPL to GPL for non-paying customers. This makes it very hard for a small business to use them in a for profit applications and surely this was their intention but again I forgave them, it must have been in their best interest.

However now that I have seen this article and can see them sawing at the branch they are sitting on, I have no idea what the top brass is thinking. It's very alarming.

Re: More Ganssle on Ada

[Micronian Coder]

When I read the article, I too felt very disappointed that Adacore gave the impression that unless you do safety critical code, you don't need to consider Ada, which is absolutely silly. If you want to avoid bugs in your software and have good maintainability, you should at least consider using Ada to help(not guaranty!) you reach that goal. This is especially true for many who don't have the sophisticated environment that the article recommends, such as small businesses, hobbyists, and students. Perhaps the message should have been "Ada is ideal for safety critical applications, but it's properties can benefit other domains where the developer wants to avoid bugs in their software and are concerned with maintainability." 


On Saturday, January 19, 2013 3:29:30 PM UTC-8, Patrick wrote:
> I am not trying to make comments about what projects Adacore or their customers are involved in. What I am infuriated about is how Adore is representing the language as a whole.
> 
> 
> 
> Let's delve further into the comment about not using Ada for a telephone application. What language comes to mind with such an application. Erlang. Hold this thought.....
> 
> 
> 
> CouchDB's primary creator made a courageous decision to rewrite his c++ application in Erlang.  Now imagine that Erickson made a statement earlier that Erlang should not be used for databases applications. Would he still have had the confidence to try?
> 
> 
> 
> Now the fear, uncertainty and doubt(FUD) floating around with Erlang right now is that it does not scale as advertised and it's concurrency model is not so great.
> 
> 
> 
> I have no idea if this is true but let's just say there was a project manager that did and that he/she went out shopping for another language. They would want a language that was had facilities for building high performance, high reliability and massively concurrent applications, sound like a language we know ?????
> 
> 
> 
> Only they read that Adacore says Ada is not a good match for this so they move on to the next language.
> 
> 
> 
> So when I started out with Ada about 14 months ago I had many assumptions that have been corrected. I thought that since it's part of GCC that I could use it for embedded design and target all the chips GCC supports. Now I understand how complex a job it is to write a runtime and even though I am disappointed it can't really be used for embedded design(as in baremetal on ARM) I forgave Adacore. They have a business to run and if writing runtimes for lots of targets is not in their best interest, they can't do it.
> 
> 
> 
> Later on I discovered that they re-licensed a large number of libraries GMGPL to GPL for non-paying customers. This makes it very hard for a small business to use them in a for profit applications and surely this was their intention but again I forgave them, it must have been in their best interest.
> 
> 
> 
> However now that I have seen this article and can see them sawing at the branch they are sitting on, I have no idea what the top brass is thinking. It's very alarming.

Re: More Ganssle on Ada

[Marc C]

On Saturday, January 19, 2013 3:17:05 PM UTC-6, sbelm...@gmail.com wrote:

> I've had the displeasure of maintaining a lot of legacy code, ostensibly
> written in Ada

I feel your pain.

I was porting a fairly large Ada 83 system where each subsystem had a typical I/O and processing flow:

- Task to monitor a socket and put messages on a queue.
- Task that implemented the incoming message queue
- Task pulling messages off the queue, processing them, and putting any new messages on a queue.
- Task implementing the outgoing queue.
- Task to monitor the queue and write messages to a socket.

Fairly pedestrian architecture, yes? Hard to mess this up, eh?

*Every rendezvous was within an accept block.*

Essentially you wound up with the World's Most Expensive Procedure Calls (tm).

By the time the port was done I had run across only two task types whose instances were actually executing concurrently with other processing.

This system is now retired.

Marc A. Criley