Re: [Bug-time] GNU Time 1.7.90 - usefulness of output

[Assaf Gordon]

Hello,

On 07/08/17 05:15 PM, Bruno Haible wrote:
> Some of the values reported by 'time' should be marked as
> "n/a" instead of 0. For example, on Linux:
> 
> $ ./time -v /tmp/clisp/build-porting32-gcc-safety0-optimized/clisp -x '(gc)'
[...]
>         Average shared text size (kbytes): 0                          -- 
> probably should be n/a
>         Average unshared data size (kbytes): 0                        -- 
> probably should be n/a
>         Average stack size (kbytes): 0                                        
> -- probably should be n/a
>         Average total size (kbytes): 0                                        
> -- probably should be n/a
[...]
> I know that the implementation of such a feature would be system dependent.
> But it's a question of credibility to not output "0" for "unknown" or "n/a".

I agree, this could be improved.
Almost all operating systems these days return zero values
in many members variables of 'struct rusage'.

But for the coming release (which I hope will be soon),
I really want to keep it exactly as it currently is (and has been for 21
years) - meaning returning "0", which are the values returned by
getrusage(2).

Later on I would be glad to explore ways to provide more accurate
information, or even OS-Dependant information (e.g. metrics that are
only available on linux, BSD, etc.).

Perhaps also compile a definitive list of 'struct rusage' members which
are not reported by each OS, and print "n/a" in them.

What do you think ?

regards,
 - assaf