Re: Help with Atomic_Components and whole array assignment

[Robert Dewar <robert_dewar@my-deja.com>]

In article <x7vsnm0vick.fsf@smaug.pushface.org>,
  Simon Wright <simon@pushface.org> wrote:

> If it were another compiler provider, I would be tempted to
> read this whole discussion as "it would cost us a bundle to
> implement this, so let's dig our heels in". I don't for one
> moment believe that that's the case here.

You miss the point entirely.

Probably most compilers DO implement MOST of what is wanted,
but MOST is not really good enough if you are trying to write
reliable portable code.

The trouble is that "it" is not properly defined, and is not
easy to define from a formal requirements point of view. Too
many people write code that they think is correct because it
happens to work using one specific version of one compiler.

It is quite interesting to note that there is MORE machine
code in the Linux kernel than there used to be, precisely
because new better optimizing versions of gcc violated
improper assumptions about code generation sequences.


Sent via Deja.com
http://www.deja.com/