27

10-May-1967 0700

Traced Karl Pingle's AROV bug as
far as the hardware. Came to the following
conclusions: It works all right if there are
no other interrupts on the system (that's only
INSTR TEST doesn't slow it, nor TS main)
but if you run a program that gets AROVF
traps and push a key on the tty, the instruction
following the one that gets the overflow will
sometimes be executed (this is true if the
following instr is a JRST on A JFCL, did not
have time to test other instrs). I planted a
patch in the system to save the PC in the case of
AR OVR traps, and by the time you get into the
APR channel, the instructions following the one that
got the overflow had been executed.

John Sauter