On 29/03/18 10:16, Jacek Caban wrote:
On 03/29/2018 04:36 PM, Zebediah Figura wrote:
For what it's worth, it seems to me that it'd be more desirable to work around a Windows bug rather than skip the test entirely.
I'm not sure I agree. If we suspect that it messes ref counting, you can't be sure that the workaround will make it behave correctly, all you know is that it happens to avoid a crash. It's likely that the library may be in a wrong state under the hood. If we hit another crash on this platform, we will need to invest more time on investigating it for questionable benefit.
Jacek
I see; that's a fair concern.