I think we have tracked a problem to the Restart not functioning correctly
in the 17C752 using hardware I2C in Master mode. The Restart feature is used
to bring the SCL and SDA lines high from the low where the transmit of the
internal address left them so a start can begin the receive part of the I2C
transaction.

The Picmaster (in 17C756 mode) performs all of the necessary starts, stops,
etc. OK, but when RSEN is asserted high after the transmit, the SCL and SDA
remain low and the subsequent Start bit pulls the lines high rather than
asserting the start. The microchip EE's on the bus could care less if there
is an intervening Stop and work fine, but many other parts get upset when
the initial transmit ends with a stop. (One of the great things about the
Microchip EE's is that I swear one could throw a 100 base T signal at them
and they would figure out what to do!)

1. Has anyone else experienced this behaviour from the ICE or the chip?

2. Suggestions on "Don't do that's" or work arounds. (The 17 series won't
let one play with PortA so bit twiddling is out.)

Tom