[PATCH:libpciaccess] Solaris support for multiple PCI segments (domains)

[John Martin]

On 03/29/11 04:14 PM, Adam Jackson wrote:

> Consider, as a counterexample, that there exist only two variables in
> the core X server that are volatile, and yet we do a very similar
> resize-the-array move all the time.  Is this because X is broken, or is
> this because your understanding of volatile is faulty?

I understand the volatile behavior as defined by the C standard.
I also work in an environment where different compilers and
compile options can be used, which spurs a bit of paranoia.
If this use of this qualifier in platform specific code is going
to cause the community sleepless nights, I can remove it.

> [1] - Unless, of course, di_walk_minor is handing that memory to
> something that runs concurrently with the pciaccess consumer, and may
> modify it asynchronously.  If that's true, though, that's one hell of a
> broken API.

This is not the case.