rawhide report: 20060716 changes

[seth vidal]

On Mon, 2006-07-17 at 15:00 +0200, Nicolas Mailhot wrote:
> Le Lun 17 juillet 2006 14:37, Jesse Keating a écrit :
> 
> > Note:  Your system will be just as unhappy if you drop power or
> > whathaveyou in the middle of a long RPM transaction.
> > In fact, that's what you are doing.
> > Yum has little to do with this.
> 
> In fact, yum has everything to do with this, and needs to be just as
> resilient to the power failure than to the ctrl+c case (or kernel oops, or
> rpm crash, or kill -9, or whatever). ie notice the next time it's run
> something went wrong and propose the user the appropriate steps to clean
> up the mess.
> 
> I don't find the "but it can happen other ways" argument very convincing.
> Quite the contrary.

We know about this problem. And I think I know what is probably the best
behavior in this case - it's a matter of doing and testing it.

The gist of it is that we write a transaction journal out that records
what is not yet finished in the yum transaction. Then if we're
interrupted we can recover.

If people have code they'd like to contribute email yum-devel list.

thanks
-sv