Re: [Aoetools-discuss] Bug#377431: [PATCH] remove bc usage

[Ed L. C. <ec...@co...>]

On Fri, Oct 13, 2006 at 01:29:00AM +0200, David Mart?nez Moreno wrote:
...
> Also, the arithmetic you are using is not really POSIX, as you are expecting 
> big numbers. The problem is that POSIX arithmetic is limited to 2147483648, 
> but not in bash, that uses a higher limit.  You probably have /bin/sh 
> pointing to /bin/bash.

There are four uses of the shell math, and only one of them can have
numbers larger than 2147483648.

  cd ~/aoetools/
  find . -type f ! -name '*~' -print0 | xargs -0 -e grep -n -e '[(][(]'
  ./aoe-mkshelf.in:12:maxslot=$((nslots - 1))

The AoE protocol has eight bits for the slot address, so that's fine.

  ./aoe-mkshelf.in:17:minor=$((nslots * shelf * n_partitions))

The maximum minor number in the 2.6 Linux kernel is 1048576, because
of the way device numbers are encoded.

  ./aoe-mkshelf.in:18:endp=$((n_partitions - 1))

Again, you can't be using so many partitions that you run out of
usable minor device numbers.

  ./aoe-stat:24:	psize=$(((512000 * sectors) / (1000 * 1000 * 1000)))

Here we have an issue.  So I guess aoe-stat is really a bash script
and not a POSIX script.

> 	Also I do not see the point in running subsequent mkshelf processes from 
> mkdevs with 'sh', as it is abash script, and it should be marked as 
> executable.

Sometimes it isn't executable, like when it's run from the aoe driver
Makefile, and it doesn't hurt to use sh.  If it was running inside a
tight loop that needed to be optimized, this would matter.

Thanks for the helpful input.  I notice aoe-stat could be made to do
the math using smaller numbers, but it is easier to make it a bash
script.  I don't think that change will be problematic for most
users.

-- 
  Ed L Cashin <ec...@co...>