Sergey Vojtovich <svoj@mariadb.org> writes:
I failed to understand why the patch would fix the assertion mentioned in the bug report. But the patch itself looks correct. That's a bit tricky... I'll explain just one of the simplest side effects of this bug, which is half way to this assertion failure. Please let me know if you want me to track it down to the assertion failure.
Thanks, no there is no need. I agree that the patch is correct and fixes a real issue. In fact, as I understand your explanation, the real problem is that the rwlocks behave incorrectly. The assertion is just how it was discovered, but it's just one symptom of a real problem. That is how I understood the patch as well. (It is often a good idea in commit messages to explain the actual problem, rather than the symptom that was used to discover the bug. Eg. I often see commit messages like "Fix a Valgrind warning in Buildbot", instead of a proper explanation about what was actually wrong and what the user-visible effect was, if any. You did explain in the commit message that the problem was incorrect load order, no need to change it I think.) - Kristian.