… that is possible fix. Or fix of one of the problems… In the old bytecode, block creation was done with a message send, so #stepToSendOrReturn never run across a block creation boundary. Now, with the block creation bytecode, if you step, it will run until the first send of access in the block, executing the block setup code, which I think is what causes the bug. 1) added #willCreateBlock that checks if the next bytecode is a block creation bc 2) #stepToSendOrReturn now checks for blocks, too. 12193 temp variable mapping wrong in the debugger https://pharo.fogbugz.com/f/cases/12193