From: Mark Biggar <mark.a.biggar@home.com>
Subject: Re: Requeue and IN OUT parameters
Date: Sat, 05 May 2001 06:27:17 GMT
Date: 2001-05-05T06:27:17+00:00 [thread overview]
Message-ID: <3AF39D26.CD232953@home.com> (raw)
In-Reply-To: 9cv3e9$gul$1@news.netmar.com
adam@irvine.com wrote:
>
> I have a couple more questions about protected objects. Here's the first
> question; the second will be in a separate post.
>
> 9.5.4(11) says that if an entry body completes other than by a
> requeue, the program returns to the caller and any copy-back of
> parameters occurs. What happens to the parameters if the entry
> body does a requeue?
If an entry body ends with a requeue the as far as the caller is
concerned the entry call isn't finished yet, so no copy back
should happen yet.
> In this example:
>
> protected PR is
> entry E1 (Count : in out integer);
> entry E2 (Count : in out integer);
> end PR;
>
> protected body PR is
> entry E1 (Count : in out integer) when Some_Condition is
> begin
> if Blah_Blah_Blah then
> Count := Count + 1;
> requeue E2;
> end if;
> Do_A_Bunch_Of_Other_Stuff;
> end E1;
> . . .
>
> In this example, does the effect of the statement "Count := Count + 1"
> become lost? Is the value E2 sees the original value of Count, the
> incremented value, or implementation-dependent?
It is implementation dependent whether "in-out" parameters are
implemented using copy-result or by-reference. So, it appears
to be implementation defined whether or not the modification to
Count is lost. Thus the above code is non-portable. Now some types
are by definition by-reference (limited, tagged, etc.) and for
these types the above is portable and the modification would
survive. Thus to make your example portable you could wrap
Count inside a limited record type and force portable behavior.
Note, one possible surprising implication of this is that if a timed
entry call with by-refernece in-out parameters times out there is
no guarantee that those parameters have not been modified.
--
Mark Biggar
mark.a.biggar@home.com
next prev parent reply other threads:[~2001-05-05 6:27 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2001-05-04 20:28 Requeue and IN OUT parameters adam
2001-05-05 6:27 ` Mark Biggar [this message]
2001-05-05 22:13 ` Robert A Duff
replies disabled
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox