Don't invert the polarity of the GPIO pins by default

[foozmeat]

This PR changes the init behavior of the MCP23008 so only the IODIR register is set for input and leaves the other registers alone. By writing \xFF to the second, IPOL, register it was inverting the polarity of the GPIO ports.

[ladyada]

i think biggest risk is whether this messes with
https://www.adafruit.com/product/1115
@caternuson @makermelissa do you have one?

[caternuson]

@ladyada I don't have a PID1115. But I do have a bare MCP23008 that I'm wiring up to test this. Stay tuned...

[makermelissa]

I don't have one of those either.

[ladyada]

@makermelissa good thing to order, get one of the RGB ones :)

[FoamyGuy]

@ladyada @caternuson would it apply similarly to the RGB versions (PIDs 1110 and 1109)? I have one of those, I'm not whether it's negative or positive. I can test it out later tonight if that is helpful.

[ladyada]

@FoamyGuy yah its the same code - either would be good for testing - either itll work or it wont :)

[makermelissa]

Ok, will add to my next order @ladyada.

[caternuson]

The real issue here is the use of a string literal in the ctor's I2C write. This expands out to having bits which get written into both the IODIR and IPOL registers:

Problem can be solved by simply adding a b in front of the literal:

device.write(b'\x00\xFF\x00\x00\x00\x00\x00\x00\x00\x00\x00')

and you get expected traffic:

I recreated issue #22 with an MCP23008 and also confirmed above change fixes it.

@foozmeat Can you update your PR to be like that? Basically identical to original code but with the additional b as shown above. The write being done in init() is using the sequential register write feature. The first value is the starting register, 0x00 in this case, not a value being sent to a register. The values that follow are then sequentially written out to the registers. So, 0xFF is the first value which gets sent to IODIR, 0x00 to IPOL, etc.

EDIT: The LCD keypads all use an MCP23017, which has a different register layout and init code and do not have this issue.

[foozmeat]

@caternuson May I suggest changing the init to writing values to named registers more like how the MCP28017 init code works? I think it reads more clearly to a less experienced person (me). I'm happy to do the work.

[caternuson]

Sure, that approach would work too. Go for it!

[FoamyGuy]

Confirmed that this change has no issues for the PID 1110 LCD. The example code from the learn guide works the same with current adafruit:master and foozmeat:master. I can check again after the discussed changes.

[foozmeat]

I set up the init method to match the MCP28017 more closely.

There appears to be a typo in

Adafruit_CircuitPython_MCP230xx/adafruit_mcp230xx/mcp23017.py

Line 68 in d244de2

self.iocon = 0x4 # turn on IRQ Pins as open drain

referencing an iocon when the property is really io_control. I can sneak that fix into this PR or do another.

[FoamyGuy]

I tested the updated version with PID 1110 LCD and it is still working as expected.

I would lean toward making a new PR for it so it's documented separately for anyone potentially looking for it in the future.

[caternuson]

Cool. Thanks. Looks good. Test it on MCP23008 and works as expected.

[caternuson]

@foozmeat Thanks for finding this and making the PR to fix it. Please open a new issue for the MCP23017 typo.