alfs Requirements [Was blank]

Bruce Dubbs bruce.dubbs at gmail.com
Mon Oct 27 15:08:50 PDT 2008


Jeremy Huntwork wrote:
> Matthew Burgess wrote:
>> On Mon, 27 Oct 2008 15:52:28 -0400, George Boudreau <georgeb at linuxfromscratch.org> wrote:
>>  
>>> One weak point in jhalfs is the progress bar. It is only a fancy
>>> timer and does not display the percent completion.  True progress would
>>> be based  on SBU calcs for each package. This idea ranked at 99 on the
>>> todo list and eventually became covered in dust. There may yet be
>>> interest in such a mod. (nudge,nudge)
>> I'd be interested, in a hopefully simple 'feature', of being able to disable the progress bar entirely through the config screens.  It's been a while since I did some tests, but disabling it certainly seemed to speed things up a fair bit for me and IIRC several other people confirmed.  See the thread at http://thread.gmane.org/gmane.linux.lfs.automated/6664/focus=6677 which has some interesting discussion on the subject including a suggestion by a certain Mr Huntwork that mirrors my own!
> 
> Such a feature already exists:
> 
> svn co svn://linuxfromscratch.org/ALFS/jhalfs/trunk jhalfs-trunk
> cd jhalfs-trunk
>  >common/progress_bar.sh
> 
> Works great. ;)
> 
> No, but really, it would be nice to have an official option, or as 
> George mentioned, to devise a new methodology. George, a long time ago 
> (you'd have to search the archives) when Manuel was first devising a 
> progress bar, I sent him one that did at least start down the path of 
> SBU calcs, etc. I think Manuel raised the objection that any calculation 
> done effectively skews the actual SBU time. Sort of like the Heisenberg 
> uncertainty principle...

You'd have to do some tests to measure the amount of time actually consumed by 
the progress bar.  It may not be that significant.  I think anything less than 
5% wouldn't really matter.

In any case, 1/10 of a second seems like overkill.  Once or twice a second 
should be fine.

   -- Bruce



More information about the alfs-discuss mailing list