Unit tests were added in  but is incomplete. Propose to provide 100% coverage for the user agent strings that Piwik is expected to parse. A large sampling of user agent strings should help in identifying deficiencies in the parser.
For example, the following is mis-identified as Firefox 3.0:
Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.5; en; rv:18.104.22.168pre) Gecko/2009041800 Camino/2.0b3pre (like Firefox/3.0.10pre)
I think in an ideal world we could aim for a 100% coverage of user agent strings.
However I am afraid that this will lead to a very complicated user agent parsing code. We need to keep this code simple (to be maintainable) and also fast (as it is executed in the Tracker).
I vote for expanding the unit tests to the 95% top user agents, as well as ensuring accurate detection of mobile devices (iphone VS ipod touch VS blackberry etc.) - if this is not already the case.
But I vote for won't fix on the 100% test coverage that would lead to a complex parsing code which is not necessary for Piwik.
Agreed: 100% test coverage for the browsers that we've already identified in UserAgentParser.php; not 100% of all known browsers in the history of the Internet.