[svlug] Structure Field Offset Measurement

[Karen Shaeffer]

On Sun, Jan 07, 2001 at 12:10:54AM -0800, J C Lawrence wrote:
> Which clearly deliniates the flaw in the thread:
> 
>   You have are attempting to discuss how a specific implementation,
>   to whit GCC on X86 when running against a Linux target, works.
>
Well that was my initial perspective. The thread has taken numerous
contorted turns along the way. I have attempted to keep to my original
perspective--but it hasn't been easy. Hahaha...

>   We have been attempting to discuss what the standard defines
>   as as required behaviour as versus what it leaves as either
>   implementation defined or as undefined (which are critically
>   different).

Agreed.

> 
> This was not clear in your initial context-less query of a macro
> definition, or in the early stages of the thread (my initial
> response stated that I was commenting on the spec, not the
> compiler).  In fact there was good evidence from you to think that
> we were discussing the same thing, then.

Agreed, But I realized folks were drawing me into an untenable position--so
it was wise to constrict my statements to the original perspective. I think
that is proper. The truth is I don't like when people jump into a thread and
change the context. It happens all the time. And once it happens, folks just
start talking past each other without stopping to think about it. Your
observation is a good one.

> What a particular compiler
> does in a well defined environment ia barely interesting at best

Hahaha. If you are interested in the Linux kernel, then you see it quite
differently. If you are interested in custom code that runs on specific
hardware and is intended to eek out the maximum possible performance--then
you see it differently. This could be quite applicable in the case of
specific applications running on Beowulf Linux clusters--as an interesting
example.

> whereas what an arbitrary compiler on an arbitrary platform platform
> when compiling for an arbitrary target will necessarily do with
> given code is a far more interesting thing -- tho my biases may be
> showing slightly here.

This is completely valid and just as interesting. It's just not my focus at
this time. Maybe someday it will be.

c,
-- 
 Karen Shaeffer
 Neuralscape; Santa Cruz, Ca. 95060
 shaeffer at neuralscape.com  http://www.neuralscape.com