maria-developers team mailing list archive

[Kristian Nielsen <knielsen@xxxxxxxxxxxxxxx>]

"Vladimir Kolesnikov" <vladimir@xxxxxxxxxxxxx> writes:

> I'm trying to reproduce this bug in PBXT:
>
> https://bugs.launchpad.net/pbxt/+bug/668491

Great!

> I use sources from mariadb-5.1.42.tar.gz 

That's quite an old version, you should use the latest.

> I got some problems trying to run maria's test suite. What I want is to run
> the pbxt.preload test. I can confirm that PBXT engine is compiled-in. I
> tried several ways like 
>
> ./mysql-test-run pbxt.preload
> ./mysql-test-run --do-test=pbxt.preload
> ./mysql-test-run --start-from=pbxt.preload
>
> but every of them failed with some kind of init error... Any ideas of a
> quick-fix?

Using source tarball should work, but I would recommend using the bzr
tree. Here is a step-by-step instruction that works for me:

    bzr branch -rtag:mariadb-5.1.53 lp:maria/5.1 work-5.1-bug66849
    cd work-5.1-bug668491/
    BUILD/compile-pentium64-valgrind-max
    (cd mysql-test && ./mtr pbxt.preload)

Note, however, that this failure is hard to re-produce. Usually when the test
is run, it succeeds. Only occasionally on the heavily loaded buildbot machines
does it fail. Here is a search against the database of past failures for this
test:

http://buildbot.askmonty.org/buildbot/reports/cross_reference#branch=5.1&revision=&platform=&dt=&bbnum=&typ=&info=&fail_name=pbxt.preload&fail_variant=&fail_info_short=&fail_info_full=&limit=100

As you can see, the failure is quite rare, only a few failures out of the
hundreds of test runs our Buildbot does each day. But according to the bug
report, I was able to repeat as follows:

    (cd mysql-test && for i in `seq 1 1000` ; do ./mtr --parallel=3 pbxt.preload pbxt.preload pbxt.preload || exit 1 ; done)

I just tried it, and I was able to reproduce with above loop, however it takes
several minutes to see the failure.

(I'm using 64-bit Ubuntu 8.04 on a 2.4 GHz Core 2 Duo laptop).

Good luck, and let me know if you need more help.

 - Kristian.