Re: Adding sha3 digests to digest

[Alistair Crooks <agc%pkgsrc.org@localhost>]

On 17 February 2016 at 12:56, Taylor R Campbell
<campbell+netbsd-tech-pkg%mumble.net@localhost> wrote:
>    Date: Wed, 17 Feb 2016 12:51:22 -0800
>    From: Alistair Crooks <agc%pkgsrc.org@localhost>
>
>    Added (via -t switch to digest). Please refresh
>    http://ftp.netbsd.org/pub/pkgsrc/misc/agc/digest-sha3-20160217.diff
>
> I suggest always calling it, unconditionally.  It's pretty cheap --
> it's the cost of hashing few kilobytes, which in the context of
> forking and execing a process isn't much.

Yeah, true. We could also ping the default gateway on every
invocation, calculate pi to a million digits, and automatically delete
trailing whitespace from all files in $CWD, before running a normal
digest, and they'd accomplish the same kind of thing as hashing 2KB of
data on every invocation of digest(1).

I'm not sure what running the self tests on every invocation would
show? The expected failure rate would be fairly small, I'd think, and
any time we did get a failure, what would it signify? How could it
help to inform someone who was motivated to fix it? Does it come from
the kernel, from mmap (unlikely, given that digest doesn't invoke
mmap(2) at all), or problems in malloc, or L2 cache coherency, or...?
i.e. what are we trying to accomplish by running the sha3 self test?