Markus Dreseler

Here's a different Hyrise result: ``` 32 11 59 19 NULL NULL NULL NULL 32 11 59 19 32 11 59 19 32 11 59 19 43 11 1 19...

`_left_row_id_has_match` is written happily from multiple threads. I'm surprised that tsan isn't red-faced. Is there any reason why that and _right_row_id_has_match are shared across different clusters? I won't find the...

"Never" is not correct. What is true is that it happens significantly more often when the test is isolated. No idea yet why that might be the case.

Pinging @julianmenzler who touched this recently.

Duplicate handling of plugins is broken: ``` (release)> load_plugin ./lib/libMvccDeletePlugin.so Plugin (MvccDeletePlugin) successfully loaded. (release)> load_plugin ./lib/libhyriseTieringPlugin.so terminate called after throwing an instance of 'std::logic_error' what(): src/lib/utils/plugin_manager.cpp:57 Loading plugin failed:...

Meta tables (at least meta_chunks, potentially more) cannot handle physically deleted chunks **PR open: #2182**

MVCCDeletePlugin physically removes the chunk but that does not cause memory to be released Update: Caused by #2178, more specifically the Delete operator not releasing the input table and thus...

plot_perf_over_time: Lines for annotations are missing if only one subplot is shown. see https://stackoverflow.com/questions/66624628/matplotlib-png-and-pdf-output-differ-lines-have-wrong-coordinates

Segment Access Counters seem to ignore attribute vector accesses

Just to document the thoughts from before: It should not be the job of indexes to handle type conversions. `TableScan::create_impl` does it for scans, other users should do it similar....