Re: Help with embedded hardware/software platform selection for ADA

[mjsilva@scriptoriumdesigns.com]

On Tuesday, July 2, 2013 3:47:17 PM UTC-7, Randy Brukardt wrote:
> <mjsilva@scriptoriumdesigns.com> wrote in message 
> 
> news:d39ad66d-63a3-43e8-a8ae-3c799c12512e@googlegroups.com...
> 
> ...
> 
> >> Why a hobbyist would even look at Ravenscar?
> 
> >>
> 
> >Because it provides the simple multitasking that a great many projects can 
> 
> >benefit from.
> 
> 
> 
> That's true only if you think hobbyists want a lot a frustration. That's 
> 
> because Ravenscar makes it impossible to build resilient systems - the lack 
> 
> of a entry call with timeout means that any error or failure will cause a 
> 
> deadlock. If you are going to use fancy proof tools to *prove* that there 
> 
> are no errors or deadlock possibilities, then this is OK (and that's what 
> 
> Ravenscar is for), but hobbyists are unlikely to have access to such tools. 
> 
> (Heck, *I* have no access to such tools - I'm not even sure if they really 
> 
> exist.)
> 
> 
> 
> Even if you don't have a sensible recovery action, the timeout can raise an 
> 
> exception pinpointing where the deadlock happened -- which greatly eases 
> 
> debugging. (I don't think Claw would have been usable without that, as it is 
> 
> way too easy to unintentially cause a deadlock by doing the wrong thing in 
> 
> an action handler.)
> 
> 
> 
> So, IMHO (and much like SPARK), Ravenscar is a tool needed only in very 
> 
> specialized circumstances (typically, some sort of certification 
> 
> requirement). Otherwise, one ought to forget about it and use Ada as it was 
> 
> intended.

Interesting point.  I understood the "why Ravenscar" to mean "why any tasking, even if limited?" which may have been a misunderstanding on my part.  Of course I'd like to get as much Ada tasking on bare metal as possible, but I'd willing to give up having the whole package to get something useable that will fit on smaller hardware.  I'd be willing to settle for some tasking sweet spot as regards code size vs tasking features.  That was the point of my reply.