OOoCon and ODF tests

["Hanssens Bart" <Bart.Hanssens@fedict.be>]

Title: OOoCon and ODF tests

Hi,


a quick round-up of the interop and testing sessions at the OpenOffice.org 2008 conference in Beijing.

The sessions have been videotaped (except for the OASIS part) and they probably be online before the end of the month.


- of course there was the OASIS ODF Interop session, featuring, among other things, an spreadsheet formula exercise. Once again, it showed us how careful we should be when performing tests.
Roughly speaking,  the test was a spreadsheet containing formulas in column A and the expected results in column B.

An older version of one implementation didn't recalculate the formulas in column A but simply reused the cell values stored by the application that created the test spreadsheet, thereby achieving a very, very high score without actually implementing formula features. The new version of this implementation, by the way, does support formulas and did perform very good on the test

Overall, formula interop was very good (at least for the configurations that were tested that day)



Back to the main OOoCon.
See also: http://marketing.openoffice.org/ooocon2008/programme.html


- there was the shared Sun-IBM announcement on the http://odftoolkit.org project, uniting the ODF validator, ODFDOM and ODF XSLT Runner. As you all remember, the OIC Charter states that the OIC itself won't be writing software, so these tools could be very useful

- DeltaXML.com offers a (non-open source) ODT compare tool which might be interesting for our purposes.

- Remember, when comparing ODF-files at the XML-level, some normalization (like c14n) has to be done. For instance, one might want to convert nested text:spans to sibling text:spans.

- Svante (Sun) quickly mentioned another tool used by Sun for regression testing: printing rendered ODF documents to an image and doing a bit-comparison with a print-out of the previous version.
While interesting for regression testing, this approach also has some drawbacks if one would like to use it for interop testing. For instance, there's no way to see the difference between a "real" table and an embedded image that looks like said table.


Best regards,

Bart