Hi Alex, while I haven't done any performance tests yet it is still my impression that Falcon is quite a bit faster, but as long as 40% of the tests don't seem to be providing useable results, it's too early to be happy about that. Don't know if the fixes to the remaining problems would slow things down.
I found some more problem Areas: - It seems to be impossible to build MXML based Modules, while AS base Modules compile nicely MXML ones don't and it doesn't matter if they are based upon Spark or MX Module. - If someone uses a CSS with styles without a namespace, Falcon dies of a NullPointerException - If I have an Application that uses a generated source class, the compiler seems to finish without complaining, but when executing the resulting SWF the flashplayer claims that for example the generated class is trying to reference an unknown function (One which is definitely part of Object). While some of the problems seem to be related with compiling really old flex 2 and flex 3 applications, some things however are real problems. I added a sub-folder to my flexmojos Falcon testsuite in which I manually created some tiny projects that reproduce the errors I identified in the official testsuite (https://github.com/chrisdutz/flexmojos/tree/feature/falcon/test/flexmojos-tests/identified-problems). I will be able to place these in the Apache repo, but I'm still working through the failled tests to find out what's going wrong. And I'll continue spotting the problems. It would be great if someone could have a look the how to fix them, doing both is a really huge task. Chris ________________________________________ Von: Justin Mclean <[email protected]> Gesendet: Donnerstag, 23. Oktober 2014 03:28 An: [email protected] Betreff: Re: Checked in failling flexmojos-with-falcon modules Hi, > While a lot of tests are failing because of unsupported compiler options such > as advancedTelemetry Advanced telemetry should be relatively easy to add (and quite useful) ,I add it to the existing compc compiler and basically it's just flag in the swf header. > or ASDoc generation. That's probably going to be a little harder, the current solution is a mess of XML and HTML templates, XML style sheets, and Java code that's not that straight forward to modified. It wouldn't be easy to port across. Out of interest is there any performance tests in there? Would be interesting to know if you getter better performance (esp on Java 1.8). Thanks, Justin
