My personal and professional life


My MySQL Bugs

Since I started using MySQL 3.23.x during my university years I was trying to solve all my problems myself like with all the other free and open source software I was using. At first it was more about learning (or lack of knowledge if you want), but then the problems become more complex and harder to solve. Well, I was successful most of the times, so I didn't consider reporting bugs or asking for support. However MySQL ecosystem grew large since then, the software become lot more complex and the source code base increased significantly, so nowadays it's native to think one could manage to fix all the problems he encounters without going deep in product's internals or relaying on professional support. The first is possible only for MySQL developers well familiar with the code base and large organizations like Facebook, Booking, etc., which have dedicated teams and regularly contribute patches. The second depends on you and of course the company providing the support.

I haven't reported much bugs for MySQL - just 21, but I now consider reporting everything I found which seems like a bug, because this is the right thing to do. I may as well solve the problem myself, but someone else could be stuck by the same and should at least be aware that it's a known (and eventually already resolved) problem. With each one of these I learned a little something, so here below I make a review of my bugs so far describing their fate in chronological order.
  1. Bug 20098 (my first one) was reported on 2006-05-26 and is with status "No Feedback" as of 2006-07-01. It was about Query browser searching for a file that is not present ( I have forgot about this one and found out it was closed later, because no feedback was provided, which is fun as I have provided the requested information...
  2. Bug 69459 was reported 2013-06-13 and is in status "Can't repeat" as of 2013-12-24. It was about MySQL Workbench crashing on entering dot in a SQL string when automatic code completion is enabled. I reported the bug for version 5.2.47 CE, but I was suggested to try 6.0.7, which worked, so the bug didn't receive more attention since 5.2 series didn't receive more updates.
  3. Bug 73076 (my first confirmed and fixed one) was reported on 2014-06-22 and is with status "Closed" as of 2014-08-26. It was about MySQL Workbench actually committing modifications in a record set when you only apply them. It was a nasty one making my daily work with Workbench difficult as I had to be careful not to apply (pardon commit) too early. The problem appeared in 6.0.x, 6.1.x and was finally fixed with the release of 6.2.2 on 2014-09-05 (see release notes) after being verified by support.
  4. Bug 73079 was reported on 2014-06-23 and is with status "Closed" as of 2014-09-15. It was about MySQL Workbench not refreshing the cell after setting value to NULL. It affected 6.1.x and early 6.2.x versions. Although it wasn't officially confirmed for me it was a regression, but more importantly it was fixed with the release of 6.2.3 on 2014-09-23 (see release notes).
  5. Bug 73708 was reported on 2014-08-25 and is with status "Closed" as of 2014-12-01. I was about MySQL Workbench models being completely mangled. I forgot this one for a few months and provided the requested feedback later to realize the problem was already fixed with the release of 6.2.4 on 2014-11-20 although nothing was mentioned in the release notes.
  6. Bug 73770 (my oldest one still open) was reported on 2014-08-29 and is still with status "Verified" as of 2018-02-05. It's a feature request for MySQL synchronizer between model and database to display the actual differences. Whoever used this functionality should know that applying changes in the model to the database could be quite annoying, because there's always some changes you do not expect and furthermore you do not understand. You only see the update script generated, but it would be nice to see the actual difference, which in some cases may not even be difference (see 90772 below).
  7. Bug 82202 was reported on 2016-07-12 and is with status "Closed" as of 2018-01-31. It was about impossibility to link Connector/ODBC 5.3.6 to the shared due to undefined symbols (e.g. my_malloc, my_free, etc.). The problem was quickly explained, but then closed as "Not A Bug". It was reopened after a Fedora maintainer reported the same problem and finally fixed with release of Connector/ODBC 5.3.10 on 2018-01-30 (see release notes).
  8. Bug 84951 (my first suggested patch) was reported on 2017-02-10 and is with status "Duplicate" as of 2017-02-12. It was about problem building MySQL Workbench 6.3.9, because of compile errors in jsonparser.cpp and jsonview.cpp for which I suggested a patch. The bug was made duplicate of 84886, which also contained a patch provided two days earlier, but it was closed on 2018-05-14 as "Won't fix", because they "no longer support 32bit systems" (!?).
  9. Bug 89608 was reported on 2018-02-09 and is still with status "Verified" as of 2018-02-13. It is a feature request about having password requirements messages in MySQL Workbench match the server configuration, so they're more useful. I provided sample code about how it could be done and even wrote a SQL function myself to test the idea.
  10. Bug 89615 was reported on 2018-02-10 and is with status "Unsupported" as of 2018-08-23. It was about failure to build MySQL Shell 1.0.10, because compilation fails for undefined vio* functions. It was related to 82202 and although confirmed it was later closed with the explanation that "MySQL Shell 1.0 is no longer supported and version 8.0 should be used instead" (!?) to which I reacted a bit harsh, but I really do not understand such answers.
  11. Bug 90619 was reported on 2018-04-25 and it is with status "Duplicate" of 79315 as of 2018-04-25. It was about MySQL Installer not offering upgrade from MySQL 5.7 to 8.0, which was explained as "not a bug", because it was "working as expected", so apparently MySQL Installer doesn't support upgrade between major releases. I hope this changes in future.
  12. Bug 90620 (my first of a series) was reported on 2018-04-25 and is still with status "Verified" as of 2018-04-26. It is about MySQL Workbench 8.0.11's SQL editor displaying error on SELECT query with window functions. You wouldn't expect MySQL Workbench having the same version as the server to lack support for new server functionalities, but more about this later.
  13. Bug 90727 was reported on 2018-05-03 and is with status "Closed" as of 2018-06-01. It was about impossibility to link Connector/C++ 1.1.11 due to missing mysql_sys and mysql_strings libraries. It turned out to be a regression from some Solaris specific modifications not made Solaris specific. I got in touch with the developer and it was declared fixed in 1.1.12, which got released on 2019-01-28 (see release notes), so for 1.1.11 I had to patch it myself.
  14. Bug 90772 was reported on 2018-05-06 (St George's Day) and is still with status "Verified" as of 2018-05-09. It is about MySQL Workbench synchronizer making difference between single quotes escapes (e.g. the statement COMMENT 'Currency\' symbol' is different from the statement COMMENT 'Currency''s symbol'). It is because MySQL Workbench allows backslash escape and the server accepts it execution, but then rewrites it with single quote escape. Thus, on the next synchronization the difference remains (e.g. like you haven't synchronized at all).
  15. Bug 90876 was reported on 2018-05-15 and is with status "Closed" as of 2018-06-21. I was about MySQL Shell 8.0.11 giving error 5115 on adding documents to collection in Server 5.7, but what it was actually about was automatically generated identifiers as I explained in details in my article MySQL Shell 8 and automatically generated document IDs. This one resulted in documentation update and since nobody else experienced this problem I guess it's fine, but for me such problems break backwards compatibility.
  16. Bug 91841 was reported on 2018-07-31 and is with status "Closed" as of 2019-01-25. It was about impossibility to build MySQL Connector/ODBC 5.3.11, because of compilation errors. I was able to fix the compilation errors myself (see patch) until waiting for help and explanation. And the explanation was that it was duplicate of an internal bug, "which explains the reasons for the failure", but which I'm not able to read. The funny thing about this one was that there were more than 400 lines wiped out from the beginning of a header most probably related to copyright text updates. It was fixed with the release of 5.3.12 on 2019-01-28 (see release notes).
  17. Bug 92898 was reported on 2018-10-23 and is with status "Not a Bug" as of 2018-10-23. I consider this as my one and only "support request" and I was pleased, that it was explained the same day.  It is about hitting an incompatible change during upgrade from 8.0.12 to 8.0.13, because I still had same old views using ASC or DESC qualifiers for GROUP BY clause. In this regards I'd like to thank MySQL for the detailed release notes they publish although as in this case even if you read them carefully you may omit something by considering it doesn't concern you. For this reason it's highly recommended to use MySQL Shell's Upgrade Checker Utility, which could tell you about compatibility errors and issues in advance.
  18. Bug 92900 was reported on 2018-10-23 and is with status "Verified" as of 2018-10-23. It is about MySQL Workbench 8.0.13 lacking support for expressions in DEFAULT although the feature is supported by the same version of the server.
  19. Bug 92908 was reported on 2018-10-23 and is with status "Verified" as of 2018-10-23. Is is about MySQL Workbench 8.0.13 lacking support for expressions as key parts although the feature is supported by the same version of the server.
  20. Bug 93835 was reported on 2019-01-07 and is with status "Verified" as of 2019-01-10. It's a feature request for having MySQL Workbench display a warning on using keyword as identifier (e.g. table and column names). I opened it after Frédéric Descamps suggested so in a comment on LinkedIn. Fingers crossed it's implemented.
  21. Bug 94012 was reported on 2019-01-23 and is with status "Verified" as of 2019-01-23. It is about MySQL Workbench 8.0.14 not recognizing the new LATERAL keyword and giving error after it in SQL editor. It's by now the last one of the series on MySQL Workbench not supporting the features of the server even though they're the same version. The bug was nominated as MySQL Bug of the Day by Valerii Kravchuk on the same day.
I hope I haven't bored you with all the information above. It's not always easy to properly explain a problem, but I have quite some experience being on the "other side", so I think I know what information is required for a problem to be investigated thoroughly. Still reporting bugs properly is probably an art and I'm looking forward to Valerii Kravchuk's presentation How to create a useful MySQL bug report at FOSDEM on Saturday.

Some of my bugs are pure technical ones (e.g. 82202, 84951, 89615, 90727 and 91841) for which I reported all the details and still the explanations on some of them seem pretty strange to me (e.g. no longer support 32bit systems, MySQL Shell 1.0 is no longer supported and version 8.0 should be used instead) even though the problem and what should be fixed was quite clear. I do not understand how could you not release a new maintenance version when there are open bugs for the last such release in the series? What vendors should understand is that as a package maintainer for me it's more complex to support patches for problems not fixed upstream.

I've made a series of bugs about unsupported sever features in MySQL Workbench (e.g. 90620, 92900, 92908 and 94012), which made me think about the meaning of version number alignment that was done with the release of MySQL 8.0.11 as General Availability release (see MySQL 8.0: It Goes to 11!). As I understand the post it was done, so that we use the "correct versions" of the all the products. It's generally a good idea, but shouldn't the "correct version" support the same feature set of the server? Users should be presumably less confused if so, but so far this is clearly not the case with Workbench. I hope the situation improves in future.

I'd like to complete these post with some final thoughts. As a developer I'm aware that could never be a software completely free of bugs. However, as a user I expect that when a bug is properly reported and the problem is clear the solution to not take that much time, but of course what really matters are priorities and they are normally as higher as the number of affects users is bigger.

No comments: