[Monetdb-developers] other optimizer pipelines

Stefan Manegold Stefan.Manegold at cwi.nl
Fri Apr 9 16:49:30 CEST 2010


please note that the bug you reported is (was) not specific to the minimal
optimizer pipeline.  Setting MonetDB environment variable sql_optimizer ---
either in the monetdb5.conf file or on the mserver5 command line to any
(correct and consistent) comma-seperated list of optimizers, rather than
using the pipeline names predefined in monetdb5.conf did trigger the crash
with the second mclient connection.

Hence, a respective test would need to start mserver5 and set the
sql_optimizer variable, either in a (local) monetdb5.conf file ("trick"), or
on the mserver5 command line (e.g., start mserver5 "by hand" in a python
script (preferred) or a shell-&-batch script pair), or by using a respective
MAL command --- I could not fine any, that does this, though.

I'll think about the details...


ps: Indeed, running all ("some" means someone would have to make a
    "reasonable" choice --- I wouldn't know where to start nad stop choosing
    ...) SQL tests through multiple (predifined) optimizer pipe-lines or
    even all(?) possible/reasonable combinations of optimizers is
    currently(?) no option, as a night is simply to short for that ...

On Fri, Apr 09, 2010 at 10:12:43AM +0200, Wouter Alink wrote:
> Dear M5 developers,
> Yesterday I reported a bug regarding the minimal optimizer pipe (ID:
> 2983773). And, as expected, I was asked to add a test for the issue,
> which I would like to do.
> The bug is triggered as follows:
> - start an mserver5 instance with the optimizer pipe set to the
> minimal pipe: "inline,remap,deadcode,multiplex"
> - connect with an mclient
> - connect again with an mclient
> This is obviously very easy to reproduce.
> But, I couldn't find any other test that sets a different optimizer
> upon startup of the mserver. I found a few tests that change the
> optimizer within a SQL script, but this will very likely not trigger
> bug (ID: 2983773).
> Then I was wondering; there are currently a few different optimizer
> pipelines defined in the monetdb5.conf. Isn't it a good idea to run at
> least a few (preferrably all) tests for each of those optimizer
> pipelines? I can imagine that testing all optimizer pipelines will
> take too much time for nightly testing, but running the testweb with
> different pipelines would probably trigger most obvious bugs that are
> currently found one-by-one. Bug #2983773 would most probably have been
> detected too.
> Anyway, I would like to add a test, and I guess testing multiple
> optimizer pipelines won't be high on the priority list. Therefore,
> could somebody perhaps point out how I could specify in the 'prologue'
> which optimizer pipeline to use? I did find a python file
> "sql/src/test/Connections/Tests/connections.py" which perhaps could be
> used? What is the preferred way for adding a test with a different
> optimizer path?
> Kind regards,
> Wouter
> ------------------------------------------------------------------------------
> Download Intel® Parallel Studio Eval
> Try the new software tools for yourself. Speed compiling, find bugs
> proactively, and fine-tune applications for parallel performance.
> See why Intel Parallel Studio got high marks during beta.
> http://p.sf.net/sfu/intel-sw-dev
> _______________________________________________
> Monetdb-developers mailing list
> Monetdb-developers at lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/monetdb-developers

| Dr. Stefan Manegold | mailto:Stefan.Manegold at cwi.nl |
| CWI,  P.O.Box 94079 | http://www.cwi.nl/~manegold/  |
| 1090 GB Amsterdam   | Tel.: +31 (20) 592-4212       |
| The Netherlands     | Fax : +31 (20) 592-4199       |

More information about the developers-list mailing list