ANN: Simple Components v3.1

ANN: Simple Components v3.1

[Dmitry A. Kazakov]

   http://www.dmitry-kazakov.de/ada/components.htm

This version is compatible with new GNAT GPL 2008.

Other new things are:

- Lock-free FIFO of indefinite objects;
- For signaled FIFOs waiting for not full or not empty queue can be now
canceled from outside;
- A test added for a GNAT GPL 2008 bug.

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

Re: ANN: Simple Components v3.1

[christoph.grein]

On 29 Jun., 21:15, "Dmitry A. Kazakov" <mail...@dmitry-kazakov.de>
wrote:
>    http://www.dmitry-kazakov.de/ada/components.htm
> ...
> - A test added for a GNAT GPL 2008 bug.

It would be interesting which bug you found.

Re: ANN: Simple Components v3.1

[Dmitry A. Kazakov]

On Sun, 29 Jun 2008 22:31:07 -0700 (PDT), christoph.grein@eurocopter.com
wrote:

> On 29 Jun., 21:15, "Dmitry A. Kazakov" <mail...@dmitry-kazakov.de>
> wrote:
>> � �http://www.dmitry-kazakov.de/ada/components.htm
>> ...
>> - A test added for a GNAT GPL 2008 bug.
> 
> It would be interesting which bug you found.

The bug is in initialization/finalization of controlled objects. It appears
when a record type is initialized by an aggregate. The record type has a
controlled component and a discriminant. The controlled component in the
aggregate is produced by a function call. The result would be missing
component finalization when the function result is copied.

This constellation is not very common, so it is not that catastrophic. My
guess is that the bug is a side effect of freshly introduced limited
aggregates/return functions. So I didn't report it separately. I already
reported another bug of premature finalization to AdaCore. Both I believe,
is just one issue.

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

Re: ANN: Simple Components v3.1

[christoph.grein]

> This constellation is not very common, so it is not that catastrophic. My
> guess is that the bug is a side effect of freshly introduced limited
> aggregates/return functions. So I didn't report it separately.

You should report it as a follow-on. My experience is that sometimes
in such cases, only the part that you reported is corrected and the
rest (which you thought is the same problem) remains.

(And if you're not a supported customer, the defect will stay
therefore for another year possibly until you find it in the next
GPLed version.)

Re: ANN: Simple Components v3.1

[Dmitry A. Kazakov]

On Mon, 30 Jun 2008 01:13:23 -0700 (PDT), christoph.grein@eurocopter.com
wrote:

>> This constellation is not very common, so it is not that catastrophic. My
>> guess is that the bug is a side effect of freshly introduced limited
>> aggregates/return functions. So I didn't report it separately.
> 
> You should report it as a follow-on. My experience is that sometimes
> in such cases, only the part that you reported is corrected and the
> rest (which you thought is the same problem) remains.
>
> (And if you're not a supported customer, the defect will stay
> therefore for another year possibly until you find it in the next
> GPLed version.)

(I am a lucky supported customer.) Presently, I am waiting for a feedback
from the GNAT 6.2 wavefront team. Then I certainly will test both.

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