Re: seperator error in __mask_snprintf_len

[Joe Korty <joe.korty@ccur.com>]

On Thu, Jan 15, 2004 at 09:14:02PM -0800, Paul Jackson wrote:
> I give, Joe.  Given the several details that are better with your
> solution, I endorse your solution, with the couple of minor edits you
> have in the pipeline.
> 
> It pains me to see the minor code growth (parsing went from 391 bytes
> of machine code to 625), with non-trivial code duplication of the
> simple_stroull() routine, and admitted increase in code complexity.
> 
> But, yes, better bits than bytes, better not to alloca(), and
> better using existing bitops than misplaced arch dependencies.

First of all, I don't like my parser anymore so I hope you don't back
out, Paul.  Perhaps all that is needed to make your parser acceptable
to Andrew is 1) tweak it to use bitmap_shift_right / set_bit, and 2)
use nbits in the interface but immediately convert it to the nbytes that
the algorithm actually wants.

Over the weekend, I may poke at my version and look over yours again
and perhaps yet a third version will come out of this.  Which is a good
thing since lots of choices to pick and merge from is what is best for
Andrew and for Linux.

Joe