On Tue, Apr 30, 2024 at 06:38:40PM +0200, Mickaël Salaün wrote:
On Tue, Apr 30, 2024 at 08:13:04AM -0700, Jakub Kicinski wrote:
On Tue, 30 Apr 2024 15:54:38 +0200 Mickaël Salaün wrote:
Jakub, can you please review it?
I looked thru it. I don't have the cycles to investigate and suggest a better approach but the sprinkling of mmaps(), if nothing else, feels a bit band-aid-y 🤷️
The only mmap that could have side effects on existing tests in the _metadata one, but in fact it would reveal issues in tests, so at the end I think it's a good thing.
I'd like "self" to not be conditionally shared but that would require changes in several tests. Let's keep that for another release. :)
I also noticed that mmap() is already used in test_harness_run() with results.
Yeah, I was initially worried about adding this complexity, but at the end of the day it actually makes things more robust. I'm in favor of it.
-Kees