[Toybox] [PATCH] Fix human_readable HR_SI.

[Rob Landley]

On 09/06/2015 02:23 AM, enh wrote:
> Fix human_readable HR_SI.

For a definition of "fix" that means "decimal is no longer the default
without a flag". So this SI standards body thing is decimal units...?
That's non-obvious to me.

> The sense of HR_SI got flipped (because the tests were all backwards).
> With this, we get the same results as coreutils in the directories
> I've tried, with the exception of cases where we give "better" results
> by rounding to nearest rather than towards infinity.
> 
> Also extend the tests a little.
> 
> Maybe we should rename HR_SI to HR_1000, since "SI" is clearly
> confusing? (coreutils uses --si to mean this, but I suspect it goes
> largely unused.)

Yeah, HR_1000 sounds like a better name.

> diff --git a/lib/lib.c b/lib/lib.c
> index 27305ec..f5ed52f 100644
> --- a/lib/lib.c
> +++ b/lib/lib.c
> @@ -870,7 +870,7 @@ void names_to_pid(char **names, int
> (*callback)(pid_t pid, char *name))
>  int human_readable(char *buf, unsigned long long num, int style)
>  {
>    unsigned long long snap = 0;
> -  int len, unit, divisor = (style&HR_SI) ? 1024 : 1000;
> +  int len, unit, divisor = (style&HR_SI) ? 1000 : 1024;

While we're at it the HR_B thing needs changing, and let's change the
name (which affects df)...


> -testing "human_readable" "test_human_readable 123456789" "123M\n" "" ""
> -testing "human_readable -b" "test_human_readable -b 123456789" "123MB\n" "" ""
> -testing "human_readable -s" "test_human_readable -s 123456789" "123 M\n" "" ""
> -testing "human_readable -i" "test_human_readable -i 5675" "5.5k\n" "" ""
> +testing "human_readable l 1024" "test_human_readable 123456789" "118M\n" "" ""
> +testing "human_readable l 1000" "test_human_readable -i 123456789"

What do the "l" and "s" mean in this context? (Previous one showed the
flags...)

> "123M\n" "" ""
> +testing "human_readable s 1024" "test_human_readable 5675" "5.5K\n" "" ""
> +testing "human_readable s 1000" "test_human_readable -i 5675" "5.6k\n" "" ""
> +
> +# An example input where we give a better result than coreutils.
> +# 267350/1024=261.08. We say 261K and coreutils says 262K.
> +testing "human_readable" "test_human_readable 267350" "261K\n" "" ""
> +
> +testing "human_readable -b" "test_human_readable -b 123456789" "118MB\n" "" ""

There was an earlier argument that HR_B meant use "B" instead of no
metric for bytes. Not _always_ add a B?

I changed it to do that, but if it's wrong...

> +testing "human_readable -s" "test_human_readable -s 123456789" "118 M\n" "" ""
> +testing "human_readable -bs" "test_human_readable -bs 123456789" "118
> MB\n" "" ""

Fun corner case: If we ask for a space but not HR_B we get a trailing
space but no units. Is that right?

Rob

 1441588175.0