cvs commit: src/share/examples/drivers make_device_driver.sh

[Yar Tikhiy]

On Thu, Aug 10, 2006 at 03:09:43PM +0400, Roman Kurakin wrote:
> Yar Tikhiy wrote:
> >On Thu, Aug 10, 2006 at 02:06:18AM +0400, Roman Kurakin wrote:
> >  
> >>Robert Watson:
> >>
> >>    
> >>>On Tue, 8 Aug 2006, Roman Kurakin wrote:
> >>>
> >>>      
> >>>>We could add a "foo" driver to the LINT generated by this script with 
> >>>>notification that it was generated and all changes should be placed 
> >>>>to script.
> >>>>        
> >>>I guess I can't help but wonder if a script is necessarily better than 
> >>>a well-commented template.
> >>>      
> >>Today I've talked with ru at . He suggests to put an example to tree and 
> >>change script to produce
> >>a copy of driver using example from tree as a template via sed or smth 
> >>like that. It sounds like
> >>a good idea. We could have a single dir in subtree with various examples 
> >>and a set of scripts that
> >>could produce from one of the examples a driver to start with.
> >>    
> >
> >...and if you manage to make the driver do some real job so that
> >people start using it in production, it will have nearly zero chance
> >to fall into oblivion; it will remain in a very good shape instead.
> >No, I'm telling nonsense here:  A useful driver will be hacked
> >sooner or later into a huge, ugly, incomprehensible beast reeking
> >with features. :-)
> >  
> If you want to add a new functionality you just need to create a new 
> example.
> Probably it wouldn't be bad to have a driver with higher level of complexity
> that can do the same (and more) than the other one that is much more simple.

Of course, it's a very good idea, as long as the complex driver
source stays comprehensible to readers.  Usually that means adding
a lot of clear comments to every new feature and around it.  OTOH,
it shouldn't be a halting task because a sort of a good comment is
a must in the commit message anyway.

-- 
Yar