On Thu, Jan 20, 2022 at 3:22 PM Nikolay Sivov nsivov@codeweavers.com wrote:
On 1/20/22 16:38, Matteo Bruni wrote:
Nice find. I guess it makes sense that they did the same thing with the state objects. Weird, but...
At least it's consistent so far.
While looking at this I realized something mostly unrelated: I guess we're supposed to free the descriptors and return NULL from GetBackingStore() after Optimize() is called? No need to do anything about it right away in any case, besides maybe extending test_effect_optimize() to show the expected behavior.
I don't know about that, I haven't tried. We'll still need this storage for updates through expressions. My impression was that Optimize() strips "nonessential" parts, like strings/annotations only.
Indeed, I didn't think about expressions. So, probably nothing to worry about.