A Win95 GNAT and UNIX GNAT incompatibility

A Win95 GNAT and UNIX GNAT incompatibility

[Raymond W. Magness]

Ladies and Gentlemen,

In the course of finishing this project, I came across a version
incompatibility between Win 95 GNAT and UNIX GNAT.  What I found is:

When you have a function that returns a pointer, UNIX GNAT does not like
it when you immediately use that pointer to access a data element in the
object being pointed to.  Example:

If the function GET_ITEM returns a pointer to an object LINE which is a
record and has a field called DATA.

Do NOT do this in UNIX GNAT (but is ok in Win95 GNAT):
        SOME_DATA := GET_ITEM.DATA;     -- you will get a GNAT Bug
Error!
Unix likes this better:
        LINE_PTR := GET_ITEM;
        SOME_DATA := LINE_PTR.DATA;

I don't know why it does this, but it cost me a couple of hours to
figure it out :(  I hope no one else has had to go through this.
Later...
-- 
Raymond W. Magness
Sergeant, United States Marine Corps
Computer Science, UCLA
magness@ucla.edu 	http://www.seas.ucla.edu/~magness

Re: A Win95 GNAT and UNIX GNAT incompatibility

[John English]

Raymond W. Magness (magness@ucla.edu) wrote:
: In the course of finishing this project, I came across a version
: incompatibility between Win 95 GNAT and UNIX GNAT.  What I found is:

... but have you reported this to report@gnat.com?

There are many curious problems like this (well, not *that* many --
stop hitting me, Robert! :-) but they are GNAT problems rather than
Ada 95 problems, and the GNAT people need to be told more urgently
than "the general public" do.

---------------------------------------------------------------
 John English              | mailto:je@brighton.ac.uk
 Senior Lecturer           | http://www.comp.it.bton.ac.uk/je
 Dept. of Computing        | fax: (+44) 1273 642405
 University of Brighton    |
---------------------------------------------------------------

Re: A Win95 GNAT and UNIX GNAT incompatibility

[Robert Dewar]

John English says

"There are many curious problems like this (well, not *that* many --
stop hitting me, Robert! :-) but they are GNAT problems rather than
Ada 95 problems, and the GNAT people need to be told more urgently
than "the general public" do."

Yes, indeed, if I see a report of something that might be a GNAT bug
posted to CLA, I completely ignore it. For one thing, it is usually
woefully incomplete, and lacks the necessary information for a bug
report (full sources, detailed account of situation, version of GNAT
being used etc.)

Some of these "curious problems" are undoubtedly GNAT bugs, some are not,
but posting them to CLA will not bring them to the attention of the GNAT
team!

P.S. the fact that you get different behavior on two versions of GNAT does
not necessarily mean that you have found a bug, there are legitimate
system dependent differences, and of coruse erroneous programs may well
produce varying results. It is impossible to distingiush such cases from
genuine bugs without full information.