If you compile the app from the command line using "lzc" does that change the behavior?
When you compile it "lzc --runtime=swf10", it will print out the name of the tmp directory it is using, can you tar up that directory contents and send it to me? On Wed, Jul 28, 2010 at 12:48 PM, Ryan Maslar <[email protected]> wrote: > I currently only have 1 file being imported. The classes exist in a > library file that's included a few lines above the import statement. Testing > the same setup in a smaller application worked beautifully. It almost seems > like maybe the timing is off or there is a cache that I haven't cleared or a > memory issue exists. Not sure. Worked fine with the nightly from a week ago. > > > On 7/28/2010 12:38 PM, Henry Minsky wrote: > > When compiling <import> libraries for swf10, the compiler does things in a > slightly different order than the other runtimes. It compiles the main app > first, so that > the classes there are emitted to a tmp directory as intermediate as3 > classes, and then it compiles each > <import> library against that directory , so that it can resolve any > references to those classes, > which will be external to the library. > > One thing it was not designed to support explicitly was defining a class in > one <import> library and then subclassing it in another library. > > Unfortunately you get subclasses being made for you all over the place very > easily now, because the compiler generates a new anonymous subclass whenever > you have an instance with a method or constraint or new attribute on it. > > Could that be what is happening? Do you have a base class defined in one > <import> library and then an instance of that class being used in another > <import> library, which might be generating an anonymous subclass? > > > > > On Wed, Jul 28, 2010 at 12:25 PM, Ryan Maslar <[email protected]>wrote: > >> That's the kicker. I built a small test case and it compiles fine. I'm >> starting to wonder if it could be memory issues. My understanding was that >> compilation is sequential. If that's the case the classes should be >> available. Hell, I would think they should be available if they're in the >> import file as well. >> >> >> On 7/28/2010 12:19 PM, Henry Minsky wrote: >> >> If you are able to build a small test case you can send me, I could try to >> track it down. >> >> If you aren't able to reduce this to get a small test case, we can try >> some other approaches. >> >> >> On Wed, Jul 28, 2010 at 12:08 PM, Ryan Maslar <[email protected]>wrote: >> >>> It happens in swf10. swf8 compiles fine. I've even tried duplicating the >>> classes in the import library file (obviously with slightly different names) >>> and it still get a class definition not found. Any way this could be a >>> memory issue? I've tried doing cache clears with no change. >>> >>> Ryan >>> >>> >>> On 7/28/2010 12:04 PM, Henry Minsky wrote: >>> >>> Which runtime are you compiling for? Does this happen in more than one >>> runtime? >>> >>> On Wed, Jul 28, 2010 at 10:54 AM, Ryan Maslar <[email protected]>wrote: >>> >>>> Ok, so I grabbed the 5.0.x nightly on the 27th as I was seeing some >>>> weird compiler issues. After installing and transferring my code over, >>>> every >>>> class instance I have in a stage=late import file says it can't find the >>>> class definition. The compiler errors look like this: >>>> >>>> org.openlaszlo.sc.CompilerError: resources/plugin_library.lzx: 132: >>>> Error: The definition of base class $lzc$class_deviceComboAct was not >>>> found, >>>> in line: dynamic class $lzc$class_m15374 extends $lzc$class_deviceComboAct >>>> { >>>> >>>> This class is defined in a compiled-in library file that is included >>>> before the import statement. 'plugin_library.lzx' is the imported file. The >>>> code in the imported file that apparently triggers the compiler error would >>>> look like this. >>>> >>>> <library> >>>> <class name="testClass"> >>>> <deviceComboAct /> >>>> </class> >>>> </library> >>>> >>>> I can't seem to replicate with a test application that is set up exactly >>>> the same way as the application having the issue. What I would like to know >>>> is if anyone has absolutely any clue why this would be an issue when it >>>> wasn't occurring in a previous nightly build of 5.0.x. >>>> >>>> As always I appreciate the help. >>>> -- >>>> Ryan Maslar >>>> >>> >>> >>> >>> -- >>> Henry Minsky >>> Software Architect >>> [email protected] >>> >>> >>> >> >> >> -- >> Henry Minsky >> Software Architect >> [email protected] >> >> >> > > > -- > Henry Minsky > Software Architect > [email protected] > > > -- Henry Minsky Software Architect [email protected]
