I did result set comparisons of patched vs. unpatched server with several grammars.
No differences were found.
It should be noted that we are unable to have lengthy runs of the randgen due to other crashes, but I did several iterations with seed=time (randomizes queries + data) and nothing was found.
The ExecutionTimeComparator was also tested (compares execution times for the same query against two servers) with the same grammars - no differences reported.
The test case no longer fails / crashes and matches MySQL. It's looking good.
I have tested this patch as well.
I did result set comparisons of patched vs. unpatched server with several grammars.
No differences were found.
It should be noted that we are unable to have lengthy runs of the randgen due to other crashes, but I did several iterations with seed=time (randomizes queries + data) and nothing was found.
The ExecutionTimeCo mparator was also tested (compares execution times for the same query against two servers) with the same grammars - no differences reported.
The test case no longer fails / crashes and matches MySQL. It's looking good.
I recommend proposing a merge / will propose one.
Thanks for the work on this, Prafulla.