The only difference is that with the new allocation method the PD functions are used, while previous versions of flext used the ones of the c library. I'm not sure if the PD memory allocation can deal with the large number of implicit allocations of the C++ language. The reason for using the PD functions are that they may be faster but even that has to be verified. I'll shortly do some tests...
i see ... but you seem to run the same function again and again ...
if bytes is larger than largealloc, you run new(bytes) which ends in an infinite recursion ;-)
oh yes, many thanks for the hint... i introduced a quick fix which is not ideal, because large allocations should still be done by the c library. I have still to do some portability tests for the cleaner general solution.
best greetings, Thomas