x86-64 on i386 (was Re: Promoting i386 version over x86_64?)

[Paul Jakma]

On Wed, 16 Dec 2009, Debarshi Ray wrote:

> He is just pointing out that there is lot more work to do than you 
> think. In other words he is contesting your claim that "The kernel 
> side apparently works fine AFAICT".

Well, I don't really know how to else to counter "there may be 
unknown bugs". Kernel sub-system interfaces generally are 
well-designed and specified (i.e. explicit widths of fields). Booting 
a system and using it for a while exercises many of the important 
ones.

Could there be bugs in some lesser-used, oddball interface? Of course 
(and I am sure there are - I think I gave an example in a thread 
earlier this year). They're likely to reasonably trivial bugs though 
(oversights in the interface specification, e.g. a 'long' instead of 
a __u32, etc). If there really are interfaces that are so messed up 
that they'd be hard to fix up, then that's probably a warning sign 
that the code may have deeper, bigger problems.

People who run into such bugs can always go back to a 32bit kernel 
(standard or PAE) until it's fixed, if it even affects them. They're 
put back in the same position as they're in now, which I'm sure must 
be acceptable.

Anyway.. I'll try look into this again later next year, and see if I 
can fix the "bugs" (in the RFE sense for yum, libvirt) I found. Was 
simply hoping to get other people interested in 32-on-64, no more or 
less.

regards,
-- 
Paul Jakma	paul at jakma.org	Key ID: 64A2FF6A
Fortune:
 	"Do you believe in intuition?"
 	"No, but I have a strange feeling that someday I will."