[PD] [pii2c] keeps and resends previous messages when addresses change

[Martin Peach]

I uploaded a new version of [pii2c] yesterday via deken. It has some
slight changes, (mainly adding optional address to the read message).
I still don't get the double sends. I attach my code for teensy that
listens on all the addresses. When I try your patch with it, the
teensy only receives on the intended address. If I do 'i2cdetect -y 1'
in the terminal it receives on all addresses 0x03-0x77.

Martin


On Wed, Sep 23, 2020 at 11:49 AM Martin Peach <chakekatzil at gmail.com> wrote:
>
> Before I test this, I notice you use Wire for the i2c slaves. Wire is
> known to not work properly in slave mode.
> I use this library for teensy i2c, it works very well:
> https://github.com/nox771/i2c_t3
> But I'll try your code later and see what I get.
>
> Martin
>
> On Wed, Sep 23, 2020 at 3:40 AM Alexandros <adrcki at gmail.com> wrote:
> >
> > Sorry for coming back to this thread after so long, didn't have a Pi at
> > hand. I have attached a very simple patch which reproduces the bug.
> > You'll need to connect two I2C slave Teensy LC to the Pi. I have also
> > attached the two Arduino files that should be loaded to the Teensies.
> >
> > What happens with this setup is that the Teensy that listens to 0x09
> > (i2c_mess2.ino file) receives messages even when something is sent to
> > 0x0A. The Teensy that listens to 0x0A (i2c_mess1.ino file) doesn't seem
> > to be receiving anything.
> >
> > One Teensy prints to the serial console and the other responds by
> > lighting up and turning off the LED on pin 13.
> >
> > On 12/9/20 11:02 μ.μ., Martin Peach wrote:
> > > On Thu, Sep 10, 2020 at 2:25 PM Alexandros <adrcki at gmail.com> wrote:
> > >> When trying to send messages to two I2C slaves from the Pi with [pii2c]
> > >> the following happens:
> > >>
> > >> If I send this message to [pii2c 0x0A]:
> > >> "write 3 127 7"
> > >>
> > >> it arrives properly at the slave address 0x0A.
> > >>
> > >> If I change the address to 0x09 with the message "addr 0x09" and then
> > >> send this message:
> > >> "write 97"
> > >>
> > >> then both messages will be sent to the respective slaves, meaning that
> > >> "3 127 7" will be sent to slave 0x0A and "97" to slave 0x09. If I then
> > >> change the message sent to 0x09 to "98", this will be sent to 0x09, but
> > >> again "3 127 7" will be sent to 0x0A. If I resend "3 127 7" to 0x0A,
> > >> "98" (the last message set to 0x09) will be sent to 0x09 as well.
> > >>
> > >> The same behavior occurs when I send the message "write 0x09 98". The
> > >> same behavior also occurs if I use two different [pii2c] objects, like this:
> > >>
> > >> [write 3 127 7(
> > >> |
> > >> [pii2c 0x0A]
> > >>
> > >> [write 97(
> > >> |
> > >> [pii2c 0x09]
> > >>
> > >>
> > > Hmmm I don't get this. Do you have a patch that will reproduce the bug?
> > > (I am also able to send messages longer than 32 bytes, with
> > > MAX_I2C_BUF_SIZE 64 in my copy of pii2c.c.)
> > > I used a teeny 3.2 as a slave (using i2c_t3 from
> > > https://github.com/nox771/i2c_t3) listening on all addresses. It only
> > > receives the messages on the addresses I set in the Pd patch.
> > >
> > > Perhaps you have a hidden [pii2c 0x0A] somewhere that is also
> > > receiving your write messages.
> > > Since [pii2c] is using the same memory space for each message any new
> > > message will effectively erase the old ones, so it's hard to explain
> > > two different messages coming from a single object.
> > >
> > > Martin
-------------- next part --------------
A non-text attachment was scrubbed...
Name: i2c_openSlave.ino
Type: application/octet-stream
Size: 3224 bytes
Desc: not available
URL: <http://lists.puredata.info/pipermail/pd-list/attachments/20200924/c5dac13c/attachment.obj>