Re: Sorry if it is not for this forum

Walter Banks wrote:
Richard Tobin wrote:
Walter Banks wrote:
"*** T. Winter" wrote:
Walter Banks writes:
Richard Tobin wrote:
[...]
Suppose a loop uses an integer variable, and the body of the
loop uses that variable as an array index. If the compiler converts
it to a pointer that gets incremented, which line does the pointer
increment correspond to?

Probably where the loop is incremented.

What if the loop is compiled to something that does not loop (and I
do *not* mean loop-unrolling). Say a series of vector instructions?

A code generator / debugger message something like
"Optimize out by code generator"

It is so dangerous not to ship code in the form it was
debugged.

I don't understand this at all. I only use a debugger when I have a
bug, and then I fix the bug. Once the bug is fixed, who cares what
form I debugged it in? If I debugged it by inspection of the code and
logical analysis (a la Richard Heathfield), should I insist my users
run it by thinking it through?

If you'd said it's dangerous to ship it in a form in which it was
not *tested*, I could see the point.

I agree. I have used debugged and tested interchangeable.

Then your original comment doesn't make sense. The context was a debugger
not being able to show the corresponding source code due to optimizations.
I take your comment in the middle to mean that "it's dangerous to ship
optimized code that you couldn't step through with a debugger". I think
that's how others took your message too, given the context. Did you mean
something different?
.