On Saturday 01 August 2015 18:48:11 Martin Steigerwald wrote: > Ok, I was not aware that you cherry-picked some optimization patches. Then > I wonder why it still feels slower than the self-compiled one. I may have > a closer look at which patches you merged.
I could not include all patches due to regressions...
> > Some changes in 1.13
> > cause unrecoverable database errors when even `akonadictl fsck` does not
> > help. Some changes break initialisation of new database.
>
> Yikes. Is / did you make upstream aware of that?
Unfortunately I had no time to properly document and report this. My priority
was to fix deadlock and number of other problems in Akonadi. Testing upstream
optimisations was not part of the agenda...
I presume upstream should pick up such issues in their own testing.
There is a possibility that some of those problems could be detected by unit
tests that we do not run...
> I accept this, Dmitry. I didn´t see any regressions with the MySQL backend
> myself, so I wasn´t aware of any. And I was using this git versions for
> weeks at least.
When I cherry-picked some optimisations I found that one or two of them broke
initialisation of new database... I think borrowing any more fixes may cause
more harm than good. My hopes are with new upstream release -- at least
testing it may be feasible...
--
Regards,
Dmitry Smirnov
---
"For every complex problem there is an answer that is clear, simple, and
wrong.
-- H. L. Mencken
signature.asc
Description: This is a digitally signed message part.
