Re: [avrdude-dev] FT2232 issues - ident yes, program no

[Hannes Weisbach]

Am 10.05.2012 um 11:49 schrieb Ing. Daniel Rozsnyó:

> Hello,
>  I am trying my new board, which got integrated a FT2232 chip to support a 
> serial port (Port B) and act as a programmer (Port A). The wiring is 
> according to avrdude.conf with a LS244 tri-state buffer.
> 
>  When I run a quick test, with chip identification, the result was ok - it 
> reads the chip ID correctly. But when it gets to actual programming, the 
> avrdude is stuck at start of programming and no other commands work (not even 
> ident), I must disconnect the USB cable.
> 
>  The command which run is:
> 
>       /usr/bin/avrdude -p atmega2560 -c 2232HIO -P usb -vvvv
> 
>  The command which wont run and gets stuck is:
> 
>       /usr/bin/avrdude -p atmega2560 -c 2232HIO -P usb -vvvv -U 
> flash:w:devel.hex
> 
>  The resulting logs are in attachment.
> 
>  Any idea where the issue is?
I believe the issue is related to the polling following a page write - the 
software reads the last byte of the page, until it is read back correctly, thus 
the page is programmed and we can continue to program the next page (or finish 
up). There are multiple problems with this code:
a) a page may not be programmed up until the last byte
b) a value 0xff is indistinguishable from un-programmed flash

I have worked on the page write code for a while now and my changes include a 
fix for both problems and I tested the code for a while now. I actually wanted 
to rework the flash read routine too and submit all changes together. However, 
I could submit a patch just fixing the write routine, if necessary.
To check, if this really is your problem you can compile avrdude with debug 
symbols and check in gdb. If I'm correct, a loop located in avrftdi.c:917 will 
never be left.

Best regards,
Hannes Weisbach