Re: [avrdude-dev] W32 code merge

[Martin Thomas]

Hello,

>Eric:
>I'm not sure if Martin is subscribed to this list. Did you also send this
to
>him personally?

I'm reading the archive from time to time and I've got Alex' mail.

-- snip --

My reply to Alex' questions:
> > So, could you please add the W32-changes to the
> > CVS? W32-sources at:
> > http://www.siwawi.arubi.uni-> kl.de/avr_projects/#avrdudew32
>
> I've been through comparing the CVS 4.3.0 code to yours and apart from
> the obvious WIN32Native differences you have added I have a couple of
> questions:
>
> 1) In par.c you have added a delaymult to effectively double the delays.
> Why did you need to do this?

Hm, I don't remember this change. Might be from some debugging test
because of problems with STK200 PPIs. The delay and sleep did not work as
expected with the original/cygwin source.
You may just remove this and test programming with a STK200 dongle. If there
are problems just send me another e-mail I will try solve them.

> 2) You have also created another usleep() and added conditional
> compilation around the usleep in ppiwin.c that I added. Why did you have
> to do this?

Yes, as already written there have been problems with the STK200-PPI.
The new usleep is based an some information from MSDN and some proposals
from Jan-Hinnerk.
The usleep in w32native.c was used for testing and did work. The usleep
in ppiwin.c has been kept in the source to keep the cygwin-build as it is.

> 3) In butterfly.c you have added some the exitprogrammer command. I
> assume this should go in also.

Yes, this has been tested. For my Butterflys this exit-programmer
"add-on" works as expected. There has been a thread in
the avrdude-mailinglist about this. You may just keep it like this, it's
just sending a character so the Butterfly's bootloader exits from
programming-
mode and jumps to the bootloader-startup-sequence where the powersaving is
handled and "jump to application" is possible. avrprog in avrstudio sends
'E' if the [exit programming] button gets pressed.

Feel free to send me a source-archive with your changes before merging the
code into CVS. I can do some tests with STK500 (mega8 and mega16),
pony-stk200 (mega16) and Butterfly if needed.

-- snip --

>Jan-Hinnerk (1):
>However, I still haven't managed to install a build-environment for
>windows, since I was quite busy for the last month (or two) ;-(
>I don't know about the others, especially Alex who has the most
>experience with Windows development.

Just install Cygwin. There is a "base-install" with preselected packages.
Additionaly the packages gcc, flex, bison, automake, autoconf and make
are needed which are not in the "base-install". They can be selected
in the Cygwin-Installer. If all packages are installed start the
shell-script
makew32.sh from the avrdudew32-archive which should build
avrdude.exe-"W32-native".

> BTW: Have you had any (positive) feedback on the native version, yet?

I've given the link to the W32-package to a some people who had problems
with avrdude-cygwin and other programming-software. I've got no
feedback from most of them. This may be seen it as "no reply is good news,
since there are no problems". Beside of this only positive feedback from
users using version 0.5 so the timing-issues should be solved.
The number of users increased with the WinAVR 4/2004 no complains
in my inbox in April.

-- snip --

about the Butterfly "exit-programmer"

>Jan-Hinnerk (2):
>IIRC, we came to the conclusion that it would be best to make the
>behaviour selectable with the "-E"-option; with sending the
>exitprogrammer-command as default.

It might be better not to include this as a command-line option at all. Just
keep
it hardcoded. The preinstalled bootloader in the AVR butterfly does not work
correctly if "exit-programmer" isn't send. The user has to toogle power or
issue a reset to activate the application. Additionaly the BF-butterloader
will
not fall in a sleep-mode if it loops inside the programming-"parser" which
is
not that good for a battery-powered device. You may look into the source
code of the BF-bootloader (gcc-port of it in the "siwawi-avr-projects"-page)


Martin