I've fixed the multicalled-sub case in 3c157b3cf0631c69ffa5aa2d55b9199bf93b22a9. The original sort-block case is trickier. Deferring the exiting-pseudo-block test is easy, but actually performing the label search at the top level of the sort block is impeded by the fact that the context stack entry doesn't include an appropriate op pointer to identify the start of the sort block. One could be added as part of the CXt_NULL behaviour, or we could invent a new CXt_SORTBLOCK type that includes such a pointer, but each of these approaches has API implications. I think they're precluded at the present phase of the development cycle. -zeframThread Previous