Repository navigation
Tags
Tags
uplm80 0.3.5 — the condition truth rule is bit 0, and thirty-one othe… …r fixes PL/M-80 tests bit 0 of a condition, not whether it is non-zero. Getting that wrong broke SDIR's hash-chain scan into an infinite loop and made PIP stamp F1, R/O and SYS onto files that did not have them. Alongside it: a PROCEDURE declared at the head of a DO block was emitted inline with nothing jumping over it, which is why ED.PLM would not compile at all; IF CARRY always read false above -O 0; constant propagation was flow-insensitive, so an ordinary counting loop miscompiled at -O 3; and -O 3 turned a = 5 into a store through address 0005H. Thirty-two fixes, 76 tests. 80un is byte-identical on its whole sample corpus; MP/M II builds 41 of 41, up from 40.
PreviousNext