sound/pcm/* bugs (was: Re: page fault panic tracked down (selwakeuppri()) - really sound/pcm/*)

Mathew Kanner mat at cnd.mcgill.ca
Sun Jan 11 10:33:38 PST 2004


On Jan 11, Cameron Grant wrote:
> >	Perhaps a heavy handed approach but until someone can untangle
> >and own this problem...
> 
> if and when real life stops getting in my way, i'll resume working on it.
> 
> >	Whimsically, I wish some super-hacker could wrestle, unify,
> >unifdef and de-kobj the sound code so I could eventually comprehend
> >it.
> 
> there really are very few ifdefs in it.  in 5.x most of them are 
> unnecessary.
> 
> de-kobjification would be absolutely stupid.  completely the wrong 
> direction.
> 
> i don't know what you mean by 'unify'.

	I'd hoped the "whimsical" beginning of my sentence conveyed
its desperate and non-literal tone.  I don't want anybody to wrestle
code.  I've seen it before and the bruises are ugly.

	But since you seem to be interested in my opinion I'll explain
a little more.

	I'm not suggested dropping kobj's from the hardware/pcm
relationship.

	I would like to see the vchans and format conversion either
dropped or integrated completely.

	I would like to unifdef anything possible.  Drop the use of
snd_mutex_*, and general just assume that the code base is at least
5.x (the code is already moving in that direction, though not
deliberately).

	The feeder/mixer/vchan concepts are difficult for me and I
believe that de-kobj'ing would benefit the next generation of sound
hacker to come.  Or maybe a few lines of comments would clear
everything up, it's hard to tell.

	Finally, the original statement was a plea for help.
Anybody?...

	--Mat

> 
> 	-cg

-- 
	We peer so suspiciously at each other that we cannot see
	that we Canadians are standing on the mountaintop of human
	wealth, freedom and privilege.
			- Pierre Elliott Trudeau


More information about the freebsd-current mailing list