Re: Blocking I/O and Ravenscar

[Maciej Sobczak <see.my.homepage@gmail.com>]

On 22 Wrz, 06:31, "Jeffrey R. Carter"
<spam.jrcarter....@acm.nospam.org> wrote:

> > Basically, Ravenscar forbids any blocking operations except waiting on
> > a protected entry or a suspension object.

> I'm not sure where you got this idea; I don't recall any restrictions on
> blocking by tasks.

Hm... true, I cannot find anything like this in the profile.

However, I have found the following:

"Besides, the compliance to the Ravenscar Profile reduces the
invocation call protocol set to the single asynchronous way."

That was about communication between nodes in the distributed system,
in "Generating Distributed High Integrity Applications from Their
Architectural Description", which was presented at Reliable Software
Technologies - Ada-Europe 2007 and printed in the proceedings.

If we conclude that Ravenscar does not prohibit tasks from blocking on
I/O, then the above statement has no foundations and nodes could
safely block on two-way synchronous messages (these would be mapped to
remote subprogram calls with [in] out parameters, which is according
to this paper excluded).

In short - can I explicitly block on two-way communication in a
Ravenscar-compliant distributed system?

--
Maciej Sobczak
http://www.msobczak.com/