Logging and bc. [Was: Re: r3353]

Dan Nicholson dbn.lists at gmail.com
Sun May 20 08:14:03 PDT 2007

On 5/20/07, Ag. D. Hatzimanikas <a.hatzim at gmail.com> wrote:
> On Σαβ, Μάι 19, at 02:03 M.Canales.es wrote:
> >
> > I don't know Perl sintax, thus I opted by bc when creating the script. I you
> > can create a patch dropping bc dependency while kepping the same
> > functionality, we will apply it.
> >
> Ok, first have a look at he format and if you like it, I will continue
> to port the test script to Jhalfs.
> Have a look.
> ========================================================================================
>                                 [binutils-pass1]
> Build time is:                                          2 minutes and 44 seconds
> Build time in seconds is:                               164
> Approximate SBU time is:                                1.000
> Disk usage before unpacking the package:                202689 KB or 197.938 MB
> Disk usage before deleting the source and build dirs:   422942 KB or 413.029 MB
> Required space to build the package:                    220253 KB or 215.091 MB
> Installed files disk usage:                             45256 KB or 44.196 MB

Looks pretty nice. A couple things to thing about.

1. The script would obviously be improved if it was done all in perl
instead of forking a ton of processes. But that would require one of
use to know perl :) It looks pretty nice, though.

2. We can be smarter about how we log the time to start out with.
Instead of just taking the output that `time' gives, it would be
better if there was a value just in seconds. You can do this by using
the TIMEFORMAT variable in bash or by doing some external timing like
I did before with perl.

Consider this.

$ export TIMEFORMAT=$'\ntime %3R sec'
$ time { sleep 2; echo foo; }

time 2.013 sec

Does anyone really care how much time the cpu spent on it anyway? All
we care about is how long we had to wait or not.


More information about the alfs-discuss mailing list