D
dbeavon
Guest
>> When two threads in parallel compile two child classes of the same super-class, the super-class r-code is written twice... I've never noticed the compiler writing an r-code that wasn't requested by the compile statement. Just to be clear, are you saying that the super-class r-code will actually be written to disk or will be simply interpreted (compiled dynamically in memory) by the AVM? I always assumed that any related/referenced classes are interpreted to the extent that they are necessary for compiling the targeted class. But I didn't think that there would be anything stored on disk, unless explicitly requested. Interpreting a super-class twice is a waste of CPU but doesn't cause serious failures. Whereas the concurrent file-write problems are more serious. Is it possible that Progress has already resolved this, and it is not longer an issue? Otherwise do you know if it is documented somewhere in the KB (that classes may only be compiled one at a time across a given server or within a given output directory)? I suppose a workaround would be to have concurrent compile operations that save their results into different output directories and then merge all the r-code files as a last step.
Continue reading...
Continue reading...