Hi Dmitri,
thank you very much for your input.
I think, this all will require a major effort. In no way this can be
completed in Q2. This means, I won't start with it this month any more.
I got a couple of questions when reading your suggestions. But I will
raise them only when I look at it again next month.
Please find more comments below.
Dmitri Lenev wrote:
> Hello Ingo!
>
> * ingo@stripped <ingo@stripped> [07/05/21 19:55]:
>> ChangeSet@stripped, 2007-05-21 17:48:22+02:00, istruewing@stripped +23 -0
>> Bug#26379 - Combination of FLUSH TABLE and REPAIR TABLE corrupts
>> a MERGE table
...
> Here is the approximate list of steps which will be required to
> do this (each step is prefixed with tentative list of functions
> affected by it):
...
> Of course, changes described above are quite intrusive so probably it
> makes sense to do them only starting from 5.1.
It might even be too late for 5.1. It's a complete refactoring of merge
table handling, much more than what I tried.
> Also it might be a good idea to do changes described in points 1-4
> and in points 5, 6 and 7 as separate patches. This should simplify
> review and make the changes more manageable.
I doubt that I will get the test suite happy without 5-7. But I'll give
it a try.
> As always I (as well as whole Runtime team) will be happy to
> provide any additional explanations. We also can lend you a hand
> if you encounter some problems with implementation.
Thank you. I will really need it. Especially because it is insufficient
to have a solution that works with current code. I need to avoid
conflicts with future developments and I don't know exactly what you plan.
Facing the probable schedule of this patch, I wonder if you can think of
a small, preliminary fix for the most urgent problems: 26867, 26377, and
26379? I think we shouldn't leave them unfixed in older versions.
Regards
Ingo
--
Ingo Strüwing, Senior Software Developer
MySQL GmbH, Radlkoferstr. 2, D-81373 München
Geschäftsführer: Kaj Arnö - HRB München 162140
