Re: [Qemu-devel] [RFC] Introduce module API to QEMU

[Avi Kivity]

Anthony Liguori wrote:
> > Hmm, this raises annother issue - we've got to be extremely careful 
> > about ordering. It's not inconcievable that the PCI support code 
> > would have constructors (e.g. to register a PCI bus type).
> >   
>
> Looks like constructor/destructor has explicit support for ordering.  
> Neat.

I'd avoid it.  Have the constructor do a *pci_module_table++ = blah and 
iterate through that later.

I prefer to have the code explicit somewhere rather than gobs of 
unportable magic.  Use it where you have to, but not elsewhere.


--
I have a truly marvellous patch that fixes the bug which this
signature is too narrow to contain.