18f4520 working modbus don't work on 18f46k22

[Ttelmah]

Yes, those are only warnings. Caused because the modbus code calls the
CRC calculation functions both inside and outside the interrupt.

Now this means the PIC code is accepting the incoming data request.
So has to be accepting this and it's checksum.
It is then replying. So serial RX and TX have to be working.
What is failing is that the PC code is saying the packet checksum is wrong.
Suspect one will have to dismantle the packet and calculate the check
oneself and see what is wrong.

[temtronic]

What's very odd is that it works using the 4520 PIC and not the 46K22. Since all the modbus stuff, like CRC, is in the 'modbus.c' file and the hardware (RS485) is the same, it's hard to understand what is causing the problem...

[Ttelmah]

Question.
Was the 'working' 4520 code, compiled with the same compiler version?.
I ask, because I don't have 5.076. Now this is potentially interesting, because
as compiler versions come along, I don't save them, when I have had
significant 'issues' with a version. The fact that I didn't save 5.076, means
that for me, it had some significant issue(s).
Looking at the data packet shown, the checksum is wrong, so the PC
is reporting correctly. The data looks completely correct except for the
checksum. So I'm going to take a little while and generate the same
packet, and try putting this through the checksum calculation on 5.078,
and see if I get the result being shown here. If not, then it implies there
is some fault in the arithmetic on this compiler version.
Will come back when I have done this.

[Ttelmah]

Looking at it, something very odd is going on with his compiler version.
If you look, the code as posted has 0x1D 05 loaded into the bottom holding
register, but the reply being shown has 00 01 being sent for this.
I've changed what is loaded to the holding registers to match what he
shows being sent, then run the code compiled with 5.078, and this generates
the CRC as D8FC, not the DF08 he has. I then tried again with 5.075, and this
too gives the right CRC. A look at the 'compiler changes' list shows that
5.076, had a bug with the -- operator on certain types of variable,
fixed in 5.077 I'd suspect the modbus code is actually being damaged by this
fault.
This was obviously 'why' I did not keep this version....
D8FC is the correct checksum for the data shown.

So the reason it is failing is that he has a compiler issue.

[yvzd]

[Ttelmah]

You can also see 'why' we often have to ask for people to give the compiler
version. You nicely did, so I 'went looking' to see if there might be a compiler
issue. Sadly there was...

Glad you have it working now.

[temtronic]

He must have updated the compiler between the 4520 working and 46k22 not working ??

[Ttelmah]

That's why I asked:

[yvzd]