Re: cpplib: Tweak conditional block skipping

[Richard Henderson <rth at redhat dot com>]

On Thu, Jul 26, 2001 at 06:54:14AM +0100, Neil Booth wrote:
> Well, we'd wrap.

Yes, and what are the implications of that?  Does the compiler fail?
If so, I'd say this is unacceptable.

> But 4 million is way over what the standard requires.

So?

> We could always increase it to 22 bits and 10 for the column :-)

2**22 _is_ 4MB.  And the standard requires at least 4k columns,
which is 12 bits, which leaves you with a measly 1MB for a line
number, which is probably exceeded by any nontrivial use of c++.

The only guaranteed unique position number I can see you using
is an off_t, which of course doesn't carry column number info,
but one could easily put together an nlgn or better lookup 
mechanism to get a line number, with the column as the remainder.

If you insist on directly encoding line and column, I can't see
you doing anything acceptible in under 28+16=44 bits.


r~