Re: Stdio.Buffer, need a way to revisit old positions in the output

> Yes, that would give one a full-blown-protect-you-from-havoc solution
> even if you don't know what you're doing with the buffer.
> I still think that that would be more overhead than desirable
> in high-efficiency buffer operations.

The main issue is that you are likely to get core dumps if the pointer
is invalid.

Also, there is no way to _really_ know from the pike level when the
buffer is going to output data to the file descriptor in the basic
set_buffered mode, which would make the feature somewhat problematic
to use in that case, unless the pointers are objects.

Anyway, creating and destroying 1M rewindkey objects takes 0.2
seconds, when compared to calling a function in the buffer 1M times
(sizeof, no arguments, via function pointer), which takes 0.04
seconds.

Since you would need to call three functions to do the same thing as
the subbuffer call (get offset, get added size, update at first
offset) the time offsets are 0.12 vs 0.20 s per 1M calls on my
machine.

-- 
Per Hedbor