Fwd: GSoC NCI Purposed Improvements

[Peter Lobsinger]

---------- Forwarded message ----------
From: Peter Lobsinger <plobsing at gmail.com>
Date: Wed, May 26, 2010 at 12:01 AM
Subject: Re: GSoC NCI Purposed Improvements
To: Geoffrey Broadwell <geoff at broadwell.org>


On Tue, May 25, 2010 at 4:00 PM, Geoffrey Broadwell <geoff at broadwell.org> wrote:
> On Tue, 2010-05-25 at 13:43 -0700, jerry gay wrote:
>> > Thats not strictly true. According the C standard, short, long, long long
>> > and char all have defined limits (in <limits.h>
>> >
>> there is no 'long long' in the C89 standard.
>
> My understanding is that we limit ourselves to C89 only because we need
> to support the Microsoft Visual compilers, and that's the highest C
> standard they fully support, and it makes for a simple line in the sand.
>
> However, 'long long' is one of the post-C89 features that Visual C++
> *does* support, because so many people asked for it.
>
> So is there any good reason not to make 'long long' an allowed exception
> to the C89 rule?

I think we're missing the point of NCI here. The C89 rule shouldn't
apply to what NCI should be able to call. Parrot should be able to
call into all libraries on the host platform. This includes those that
don't arbitrarily restrict themselves to any kind of standard, let
alone C89.

Ideally, NCI signatures will be capable of supporting all legal
function signatures on all platforms (there may be pragmatic reasons
for not going so far). What this means is that NCI will support
signatures that aren't supportable on some platforms. C89 should be no
different from other platforms in this respect.