Re: Assembler in Ada?

["news.oxy.com" <Vladimir_Olensky@oxy.com>]

First of all I am glad to get response from Robert Dewar.
I would like him to be involved into some more useful discussion threads
compared with the "bracket wars" and some other small issues that usually
are internal issues of any particular organization/company.
Below are some of my comments to his reply.


dewar@gnat.com wrote in message <78fv79$7mj$1@nnrp1.dejanews.com>...


>The only time we duplicate information in the GCC manual
>(which serious GNAT users should have a copy of) is for
>commonly used features. We do not consider assembly
>language insertions to be in this category.


This passage can be easily understood  that the public distribution of GNAT
is not intended for serious GNAT users, as it does not include GCC manual.
But everything in your hands. Just include this manual  into public GNAT
distribution and then every serious GNAT user will have that manual at hand.
In my previous message I just wanted to point out that that the GNAT
documentation set is insufficient:
"ALL THE REFERENCIES FROM ANY PIECE OF THE SYSTEM
DOCUMENTATION SHOULD BE RESOLVED WITHIN THE SET OF MANUALS THAT COMES WITH
THE GIVEN SYSTEM. Otherwise the documentation does not work."

>We definitely try NOT to duplicate documentation of this
>type that is available in the gcc manual. There are for
>example dozens of switches described in the gcc manual that
>are of course relevant for GNAT, but we do not duplicate
>the description in the GNAT manual.

No need to rewrite or duplicate GCC manual. If you have references to the
GCC manual it should come with GNAT documentation set especially taking
into account that GNAT is built on top of GCC.
Se above mentioned rule.

>In addition, this feature in GNAT is powerful but you
>really need to understand quite a bit about it to use it
>on a given machine, and the understanding comes from the
>gcc manual.


Again , include GCC manual or part of it into the GNAT distribution.

>
>And yes, to use GNAT at this level, you definitely need to
>learn how to use gcc (though not necessarily to port it!)


And once more, include GCC manual into the GNAT distribution !
It is so obvious and simple !

>It would be possible to give more examples, but they would
>probably not be as illuminating as you think, because they
>would give a very limited view of what is in fact a very
>powerful facility.


No more is needed.
To my point of view several small examples and explanations in GNAT user
guide would be very useful. The idea is not to teach users deeply how to use
Assembler with GNAT but give them some initial understanding of how to use
this feature and clear understanding of what is written in Machine Code
Insertion section of GNAT reference manual. Those who want to know more will
then study GCC manual (if it will be included in GNAT distribution).
Many lectures delivered in Universities begin with the simple sentence:
"Let's consider the following example". Then you can hear some brief
explanations and then the theory behind that example.
This is the easiest way to make complex things to be felt very simple.
I hope that one of the ACT targets is make GNAT easier to understand and
work with.
Personally, I have very high opinion about GNAT and approach taken to GNAT
open source design. I would like that ACT people would make GNAT better and
better with each new release.
And do not take critical opinions of the GNAT weak sides as a personal or
team  offence. They are suggestions for improvement. Just make GNAT better
taking them into account. Sometimes such statements should more polemic to
cause some kind of reaction from the other side.

Sometimes people deeply involved into some system development (especially
software systems) have all the time to think about very specific issues so
they have less time to think more broader and generally and as a result some
obvious things may slip their minds.
Greatest example of such is the death of the one of the first US nuclear
submarine in deep waters at the yearly of the 60's.
Nuclear reactor has stopped for some reason. Crew was not able to start it
again and tried to rise submarine to the ocean surface but could not do this
because it was not possible to free ballast tanks from water using highly
compressed air.
Designers of this extremely complex submarine forgot the fundamental physics
law that is studied in the school. When compressed gas is letting out from
gas tank it's temperature is going down. Temperature difference is the
function of pressure difference. As a result of this law air pipes to water
tanks were frozen (as any gas contains some amount of water) when the crew
tried to push the water out of ballast tanks and several attempts were
unsuccessful. Submarine was going down into the deeps of the ocean and was
crashed by the water pressure. About 150 people have died in a part of the
second.


Hope to discuss more general things later. As a matter of fact this one
(assembler , not nuclear submarine)  was not very big issue. There are some
others that are of much great importance for Ada future.


Best regards,

Vladimir Olensky
(vladimir_olensky@yahoo.com)
(Vladimir_Olensky@oxy.com)
Telecommunication specialist,
Occidental C.I.S. Service, Inc. ( www.oxy.com )
Moscow,
Russia.