On Thu, Oct 15, 2009 at 06:45:47PM +0300, Michael Widenius wrote:
Hi!
"Henrik" == Henrik Ingo <henrik.ingo@avoinelama.fi> writes:
Henrik> So now would probably be a good time to blog something about our Henrik> MariaDB release. As agreed, I will blog something about how the Henrik> release came to be, Monty's will be more announcement like.
Henrik> Is the below somewhat close to what really happened?
<cut>
Henrik> <h2>Just one final merge...</h2>
Henrik> One innocent bullet point on the todo list was to merge in the newest Henrik> MySQL version, as well as XtraDB and PBXT. This was not a one day Henrik> job...
Henrik> Somehow we managed to start the merge from MySQL from a more or less Henrik> random point in the <a Henrik> href="https://launchpad.net/mysql-server/5.1">MySQL 5.1 Launchpad Henrik> tree</a>. This led to a long list of merge failures that needed to be Henrik> fixed. Quite soon we realized that we were merging something that was Henrik> between MySQL 5.1.37 and 5.1.38, and the missing changesets to 5.1.38 Henrik> were then added.
Don't think that the merge features was because we used a mid version. The problem was that we merged a version of MySQL that was in the middle between 5.1.37 and 5.1.38 and we started to chase bugs that was already fixed by MySQL (The problem is that the MySQL tree is from time to time not as bug free as it should be)
Henrik> I'm told the delta between MySQL and MariaDB is already 11kB, a lot of Henrik> it is in whitespace or comments. A good target for cleanup work in our Henrik> next release cycle...
I don't believe it. First, opt_table_elimination.cc alone is 56K. Maybe one is not running "diff -N", but even in that case the diff is going to be much larger. sql_select.cc has plenty of differences due to Maria being used for temporary tables, there is pool of threads feature. There are also a number of cleanups/improvements that were made in maria tree (and so we got them with maria engine) but haven't made it to the mainline. The diff is at least 10 times bigger than stated.
Actually, if you count the changes in the storage engines, the delta is much larger. In addition we have storage engines that doesn't exist in MySQL.
<cut>
Henrik> All in all we spent at least 4 man weeks just fixing merge issues. But Henrik> we are learning from our mistakes, and most importantly, <a Henrik> href="http://askmonty.org/wiki/index.php/MariaDB::MergingMySQL">documented Henrik> them</a>.
Great that this document was done (didn't know about it).
<cut>
Henrik> Of course Buildbot runs did report all kinds of test-suite errors, so Henrik> another batch of man weeks was spent fixing various test failures. Henrik> Many test failures happened on Windows only. To our surprise they also Henrik> happened in the stock MySQL 5.1.38 (which is GA quality). We were
Which is supposed to be of GA quality ? :)
Henrik> however able to find patches for most of those, until we were down to
find -> create ?
Otherwise looks ok to me.
BR Sergey -- Sergey Petrunia, Software Developer Monty Program AB, http://askmonty.org Blog: http://s.petrunia.net/blog