Re: Testing 3rd party engines

[Stewart Smith]

On Mon, 2006-11-27 at 23:54 +0100, Paul McCullagh wrote:
> I am just writing a blog on the subject of testing 3rd party engines  
> (http://pbxt.blogspot.com/), and was wondering if this subject has  
> come up at all before.
> 
> When using mysql-test-run to test a different engine, for example:
> 
> $ ./mysql-test-run --mysqld=--default-storage-engine=pbxt
> 
> you get a lot of failures which are not actually bugs.
> 
> For example wherever SHOW CREATE TABLE is used in a test, the result  
> verifies that ENGINE=MyISAM.

In this case we need something like the # in query results to not care
about the engine (unless we really do - e.g. explicitly setting the
engine in CREATE TABLE and expecting some behaviour).

I would consider each instance of this to be a bug (except if what i
discuss below is true and the one-true-way of doing things)

If you wanted to be really high up on the list of "people with the
number of closed bugs" you could file each failing test case as its own
bug - If it's 187 as you say in your blog post, you'd end up being just
above Monty :)


> In my blog I mentioned this and other things that would make testing  
> 3rd party engines easier by eliminating unnecessary differences.

IIRC Jeb (or someone working with him - have CC'd Jeb) worked on
splitting up the tests to be like this:

a foobar.t file containting the test

a ndb_foobar.t file that:
SET DEFAULT_ENGINE=NDB;
-- include foobar.t

(and ENGINE_foobar.t for each engine).

so you could go through and grab every test and make a small test file
for your engine. This means results are in a different result file too.

This was done to be able to run as many of the tests as possible with
NDB.


On the subject of geometry types, Brian made a patch recently that
stores the geometry data as a blob in engines that don't explicitly
support geometry types - enabling all engines to store geometry types
(not index them though).

Jeb should be able to fill you in on the ORDER BY stuff - I thought we
had gotten a smarter compare routine - obviously not :)

EXPLAIN differences probably work with the ENGINE_foobar.t test way.
-- 
Stewart Smith, Software Engineer
MySQL AB, www.mysql.com
Office: [phone] Ext: 6616
VoIP: 6616@stripped
Mobile: [phone]

Jumpstart your cluster:
http://www.mysql.com/consulting/packaged/cluster.html

Attachment: [application/pgp-signature] This is a digitally signed message part signature.asc

Attachment: [application/pgp-signature]