Avi Kivity wrote:
On 12/07/2009 06:20 PM, Anthony Liguori wrote:
Avi Kivity wrote:
Covering every qemu_malloc instance this close to the GA is too
risky. I agree that having separate behavior is less than ideal
but I think it's the only sane way forward.
I don't understand why. What's so insane about Markus' patch?
Allowing size=0 for both developer and production builds?
There is a bug here. Callers are calling qemu_malloc incorrectly.
There is an open discussion about how to address it--fix all callers
of qemu_malloc() or allow size=0. Since there isn't agreement, a
compromise of sticking to the current behavior for the development
tree, and using the later for production since we can't guarantee the
former seems reasonable.
If we apply the patch, the callers are no longer incorrect. Since
we're winding down development on that tree, I see no reason for the
production build to be correct and the development tree to be incorrect.
The problem with this whole discussion is taking the position that one
form is "correct" while the other form is "incorrect". It's not so
black and white.
I don't think a reasonable person can claim that either form of
qemu_malloc() is absolutely correct while the other is incorrect. You
may think one is better than the other but clearly, that sentiment is
not universally shared.
If we change the production build, we eliminate the immediate problem.
For 0.13, we can reduce the usage of qemu_malloc() such that it stops
becoming an issue. Then no one ever has to agree about which form is
better and we can move on with our lives.
Regards,
Anthony Liguori
