We need to parse all the warnings to decide whether they are truly duplicates or not. I can see three scenarios that we need to filter through:
- If we have two seperate warnings, with the two exact same back traces, that would be a duplicate warning we want to remove.
- Imagine we have two seperate warnings where they both have the same parent function (e.g two seperate warnings has parent function as__wt_evict_page_cache_bytes_decr) but have a totally different stacktrace. This can happen if we have two possible callers of the parent function. In this case, I would count it as the duplicate warnings. In this case the fix would be the same.
- We have two seperate warnings with different backtraces. In this case we want to treat seperately.
Defintion of Done
Able to filter the three scenarios in the tsan_warnings_analysis.py file
- is related to
-
WT-15730 test/format (disagg.mode=leader) assertion __wti_rec_row_int - Not propagating the original fast-truncate information
-
- Closed
-
-
WT-15806 Verify fails in test_layered33 (cannot find the stable on a follower)
-
- Closed
-
-
WT-15831 Fix code accidentally removed in disagg merge
-
- Closed
-
-
WT-15562 Suppress all pre-disagg warnings
-
- Closed
-
-
WT-15589 Verify each read we do is behind the page materialisation frontier
-
- Closed
-
-
WT-15750 Enable verify in test/format disagg if deltas are disabled
-
- Closed
-
-
WT-15820 test/format (disagg.mode=follower) PALite UBSAN undefined symbol error
-
- Closed
-