Re: [avrdude-dev] some problems i found, STK600 will use different pp_controlstack in some parts

[Joerg Wunsch]

As SimonQian wrote:

> 1.While programming thru JTAG as AVRDragon or JTAGICE mkII,
> AVRStudio will not send JTAGICE_CMND_RESET and
> JTAGICE_CMND_GO. These 2 commands are used when debugging thru
> JTAG. But avrdude will send these commands even when programming
> thru JTAG.

Supposedly, AVR Studio will at least send a CMND_RESET after asking
the user if the JTAG initialization fails, so the ICE can pull the
nSRST line.

Anyway, the question here is: does it hurt that avrdude forces a
reset/go?  At the very least, it provides the equivalent of what
always happens in ISP mode (because ISP requires a target reset).

Things are, of course, different for a debugger, so with AVaRICE, you
can get away without a full target reset (option -C).

> 2.I tested avrdude 5.5 and 5.3.1 and found programming speed is
> faster under avrdude 5.3.1. I will make some futher tests.

Yes, please do.

> 3.in avrdude.conf, every chip has only 1 pp_controlstack
> section. pp_controlstack is related to hardware, in the source code
> of AVRminiProg, there is a description of the usage of
> pp_controlstack. This will not trouble before supporting STK600, I
> think. For example, according to the description file of ATtiny461
> in AVRStudio, STK600 will use an other pp_controlstack, so it has 2
> different pp_controlstacks(for STK500_2 and for STK600).

The current format of the avrdude.conf file is "historically grown",
and a redesign is overdue.  Redesigning it in a hierarchical structure
is likely what would make most sense, so it's quite possible that the
"new config format" would be some kind of XML...  We've been discussing
that in the past, but without reaching a conclusion so far.

For the time being, we'll probably end up with something like
pp_controlstack_stk600 as an additional attribute. :-(

> 4.unused bits in efuse will be set to 0, which means programmed,
> while in AVRStudio they will be set to 1. I think set them to 1 is
> safer.

https://savannah.nongnu.org/bugs/?22248

I've asked the other developers here for opinions, and so far only got
one by Limor which indicates to remove the masking when reading, so it
will (implicitly) display as 1.

-- 
cheers, J"org               .-.-.   --... ...--   -.. .  DL8DTL

http://www.sax.de/~joerg/                        NIC: JW11-RIPE
Never trust an operating system you don't have sources for. ;-)