-
Type: Task
-
Resolution: Unresolved
-
Priority: Major - P3
-
None
-
Affects Version/s: None
-
Component/s: None
-
StorEng - Refinement Pipeline
Following a discussion on slack regarding the questionable barrier in __wt_yield, I noticed a similar full barrier in __wt_sleep. The comment says that "it is reasonable for someone to expect" a barrier when sleeping even if not guaranteed. I think we should instead make it explicit in all callers that need it rather than hiding it in the sleep, for the same reasons we are want to do that for yield. We can start by mechanically just adding a barrier to every call site, then remove the ones that are obviously unnecessary.
- related to
-
WT-11940 Review the use of __wt_yield (sched_yield)
- Backlog