Re: [ros-dev] [ros-diffs] [hbelusca] 58110: while (TRUE); (when something is unimplemented) ---> ASSERT(FALSE); // while (TRUE); (unless we deal with a 'noreturn' function), and in some cases, return an adequate value. Part...

[Aleksey Bragin]

this while (TRUE); means that "OS can't function anymore if this 
codepath is executed". It's a fatal, terminal way, so there is no power 
to save :)

On 04.01.2013 18:11, Javier Agustìn Fernàndez Arroyo wrote:
> well, IMO, "while(true)" means CPU is permanently busy, while an 
> ASSERT stops CPU from running ....
> just a matter of power saving....
>
> and i repeat, its just an opinion....
>
> On Fri, Jan 4, 2013 at 3:02 PM, Aleksey Bragin <[e-mail] 
> mailto:[e-mail]> wrote:
> With all respect, I don't understand many of these changes.
> Answering between the lines.
>
> On 04.01.2013 15:47, [e-mail]
> <mailto:[e-mail]> wrote:
>
>       NTSTATUS
>     @@ -643,7 +643,8 @@
>           /* FIXME: TODO */
>           DPRINT1("You have implemented the KD routines for
>     searching PCI debugger"
>                   "devices, but you have forgotten to implement
>     this routine\n");
>     -    while (TRUE);
>     +    UNIMPLEMENTED;
>     +    ASSERT(FALSE); // while (TRUE);
>       }
>
> It already prints a mandatory log message that this part is
> unimplemented. And execution is supposed to stop after printing
> this message, because it's meaningless to continue (that's why
> while(TRUE); was put there in the first place).
>
>         static ULONG NTAPI
>     @@ -678,7 +679,7 @@
>           {
>               /* /PCILOCK is not yet supported */
>               UNIMPLEMENTED;
>     -        while (TRUE);
>     +        ASSERT(FALSE); // while (TRUE);
>           }
>       #endif
>           /* Now create the correct resource list based on the
>     supported bus ranges */
>
> It already has UNIMPLEMENTED; and now you added an ASSERT(FALSE);
> which essentially is the same thing. And if continuation is
> possible, then just UNIMPLEMENTED would be enough.
>
> I'd like some general solution to this. Like,
> UNIMPLEMENTED_FATAL() or something like that.
>
> Any thoughts?
>
> Regards,
> Aleksey Bragin