Sunday, March 2, 2014

Fun with Bugs #30 - quick review of my reports in February, 2014

I've got only one comment to my previous post about deadlock, and it was more like a hint based on a different use case, not a real explanation. So far there is nobody who wants to get free beer... Maybe this is even good, as I do not go to the conference and BOF I've submitted will be supervised by my colleague Przemysław Malkowski. But you still have entire month till the conference to get a chance for a beer from him (we'll arrange this somehow).

In the meantime I'd like to review bug reports for MySQL server (few) and fine manual (many) that I've submitted in February, 2014. 22 in total, one was just plain wrong and I closed it as "Not a bug" almost immediately. So, 21 to consider.

Let's start with serious problem. Based on our previous discussion on responsible bug reporting I had not shared the test case with the entire world immediately, but you should know that you probably can crash MySQL 5.5, 5.6 and 5.7 (at least on Windows, but I can not exclude older 5.5 on Linux also) with a single SELECT statement. This was reported as Bug #71858 (security bug from the very beginning, so you will not see it directly).

While working on a real customer problem I had to report a bug related to replication, Bug #71732, "Garbage value in output when MASTER_LOG_FILE='' is set". Having this kind of messages in the error log does not really help to find out fast what was the reason of the problem.





I've considered Bug #71818, "Type column in SHOW PERFORMANCE_SCHEMA STATUS is redundant" annoying at least, but it ended up as "Not a bug" for a reason that SHOW ENGINE INNODB STATUS also has "Type" column for its single row of results, so it's by design, "The column might not be very useful, but it works as intended".

I've spent some time trying to build recent MySQL 5.5+ on Linux using Solaris Studio 12.3 compiler, and mostly failed. In the meantime Bug #71847 was reported, "Predefined compiler flags for SunPro/Solaris Studio 12.x are missing in cmake". Using section from Solaris OS part helped a bit in the process.

I was annoying enough to complain about warnings two times. See  Bug #71640, "Somewhat misleading deprecation warning for innodb_lock_monitor" and Bug #71817, "Warnings text is somewhat misleading (mentions max_open_files and table_cache)" for the details.

Surely I don't forget PERFORMANCE_SCHEMA and added a feature request, Bug #71755, "Provide per partition summary information in PERFORMANCE_SCHEMA".

The rest were actually documentation requests. I plan to write a long post about MySQL documentation in general eventually, so let me point out only bug reports I've created while working on explanation for that deadlock:
  • Bug #71637 - "Manual mentions IS_GAP and IX_GAP locks, but never explains them"
  • Bug #71638 - "Manual does NOT explain "insert intention" lock mode properly"
  • Bug #71735 - "Manual does not explain locks set by SELECT ... FOR UPDATE properly"
  • Bug #71736 - "Manual does not explain locks set by UPDATE properly"
They are all "Verified" for now, so it seems my concerns are valid.

February was a really hot and sad month here in Ukraine, so I was not productive enough with MySQL bugs reporting. Sorry.

2 comments:

  1. On behalf of the MySQL documentation team, happy birthday to you! :-)
    Stefan

    ReplyDelete
    Replies
    1. Thank you, Stefan!

      When I think about any other MySQL-related "dream job" after all these years, I am clearly dreaming about becoming a member of MySQL Documentation Team!

      Delete