This fixes EEPROM access using the STK500V2 programmer, partially
undoing part of a previous general fixup commit. Choose the correct
read/write operations with the stk500v2 program function - the correct
one depends on the memory type. EEPROM is byte addressable so uses
read/write. FLASH is word addressable and so uses read_lo/write_lo.
git-svn-id: svn://svn.savannah.nongnu.org/avrdude/trunk/avrdude@501 81a1dc3b-b13d-400b-aceb-764788c761c2
doc/avrdude.texi: (Ditto.)
Closes bug #13501: <memtype> should be listed in the man page
git-svn-id: svn://svn.savannah.nongnu.org/avrdude/trunk/avrdude@500 81a1dc3b-b13d-400b-aceb-764788c761c2
gratitous API change in recent versions of texi2html where
the output directory has changed names.
Fix for:
bug #13026: The build fails with texi2html 1.76
bug #12715: make issues during install
patch #3091: commandline fix for latest version of texi2html
git-svn-id: svn://svn.savannah.nongnu.org/avrdude/trunk/avrdude@499 81a1dc3b-b13d-400b-aceb-764788c761c2
synchronizing against a JTAG ICE in weird state.
git-svn-id: svn://svn.savannah.nongnu.org/avrdude/trunk/avrdude@497 81a1dc3b-b13d-400b-aceb-764788c761c2
that way. ;-) For those who want the 19200 Bd one, add "jtag2slow".
git-svn-id: svn://svn.savannah.nongnu.org/avrdude/trunk/avrdude@495 81a1dc3b-b13d-400b-aceb-764788c761c2
to 'i' (input). Even though this bit should be ignored, it should not
be changed. The 'x' setting sets the bit to zero which programs it
and could cause undefined behaviour. Setting to 'i' enables it to be
rewritten to its old value.
A better solution might be to read the fuse byte, apply the new value
while leaving the 'x' bit alone, then writing the value back. The
current fix is a workaround which allows the developer to change the
bit.
git-svn-id: svn://svn.savannah.nongnu.org/avrdude/trunk/avrdude@494 81a1dc3b-b13d-400b-aceb-764788c761c2
when passing unsigned char * when char * is in the prototype and vice
versa. Clean these up along with a few others.
git-svn-id: svn://svn.savannah.nongnu.org/avrdude/trunk/avrdude@491 81a1dc3b-b13d-400b-aceb-764788c761c2
#13693, #13871, and #14212.
This provides bug fixes to the STK500V2 programmer type. From the
patch information:
- incorrect token used from avrdude.conf.in
- wrong command sent to programmer, hence no write to eeprom.
- programmer was said to start writing at 0x0000 and continue page
by page and was not repositionned when a gap was found in the
hex file, or when the hex file start address was not
0x0000. Hence the verify procedure was correct, not the write
procedure.
- speed up of flash write to skip empty pages (full of 0xFF) by
re-enabling a dedicated function for that task.
- stk500v2_paged_load() was not returning the number of byte read,
so empty hex files were generated when reading memory.
Submitted by: Bernard Fouch <bernard.fouche@kuantic.com>
git-svn-id: svn://svn.savannah.nongnu.org/avrdude/trunk/avrdude@490 81a1dc3b-b13d-400b-aceb-764788c761c2
appears to have a complete XML description right now.
Document all the recently added new devices: AT90PWM2/3,
ATmega164/324/644, ATmega329x/649x, ATtiny25/45/85.
git-svn-id: svn://svn.savannah.nongnu.org/avrdude/trunk/avrdude@489 81a1dc3b-b13d-400b-aceb-764788c761c2
all the details right now, so some of the parameters are guessed.
git-svn-id: svn://svn.savannah.nongnu.org/avrdude/trunk/avrdude@487 81a1dc3b-b13d-400b-aceb-764788c761c2
changes to butterfly.c include:
. do not exit for unsupported devices but return -1 from the init function
instead; that way the -F option can be used to continue anyway
. honor the -b option as arbitrary bootloaders could be implemented with
any baud rate, not just the fixed 19200 Bd used by the butterfly
. implement functionality to read the fuse and lock bits, and write the
(boot) lock bits, resp.
. fix the signature byte order
The remaining files document the new functionality.
git-svn-id: svn://svn.savannah.nongnu.org/avrdude/trunk/avrdude@486 81a1dc3b-b13d-400b-aceb-764788c761c2
main_exit: label to ensure the programmer is released correctly.
git-svn-id: svn://svn.savannah.nongnu.org/avrdude/trunk/avrdude@483 81a1dc3b-b13d-400b-aceb-764788c761c2
thus it was falsely reporting that it failed when it was actually
working correctly. Fixed.
git-svn-id: svn://svn.savannah.nongnu.org/avrdude/trunk/avrdude@482 81a1dc3b-b13d-400b-aceb-764788c761c2
to have problems sending a control message (returns an "I/O error").
At least try to recover gracefully in the bening case where the user
did not request a particular serial number, so we could continue
anyway without knowing it.
git-svn-id: svn://svn.savannah.nongnu.org/avrdude/trunk/avrdude@481 81a1dc3b-b13d-400b-aceb-764788c761c2
usb_get_string_simple() as the latter is only found in recent enough
versions of libusb. That way, silently build without USB support
unless a recent version is available.
git-svn-id: svn://svn.savannah.nongnu.org/avrdude/trunk/avrdude@480 81a1dc3b-b13d-400b-aceb-764788c761c2
available in recent versions of libusb, and the check isn't really
needed anyway (as the check for vendor and product ID will cover that
as well).
git-svn-id: svn://svn.savannah.nongnu.org/avrdude/trunk/avrdude@479 81a1dc3b-b13d-400b-aceb-764788c761c2
mkII.
The serial transport methods have been moved out into a record of
function pointers for that purpose, defaulting to the actual serial
connection that natively applies to the hosting system. Iff inside
the JTAG ICE mkII handler a port name starting with "usb" has been
detected, the record of function pointers is switched to USB.
Optionally, a serial number might be specified, so only the JTAG ICE
mkII matching the given serial number will be opened. The match is
done right-to-left, so only the least significant bytes of the serial
number need to be given.
In order to make the change as least intrusive to existing drivers as
possible, the entire naming scheme of the serial_foo() function entry
points has been maintained as access macros that encapsulate these
into the respective indirect function calls via serdev->foo().
git-svn-id: svn://svn.savannah.nongnu.org/avrdude/trunk/avrdude@478 81a1dc3b-b13d-400b-aceb-764788c761c2
Iff enter_progmode failed with RSP_ILLEGAL_JTAG_ID, give the user a
hint that their JTAGEN fuse might be unset.
git-svn-id: svn://svn.savannah.nongnu.org/avrdude/trunk/avrdude@476 81a1dc3b-b13d-400b-aceb-764788c761c2
Now loading flash works on these devices even for simple parallel ISP
adapters.
git-svn-id: svn://svn.savannah.nongnu.org/avrdude/trunk/avrdude@474 81a1dc3b-b13d-400b-aceb-764788c761c2
of the packet to occasionally been misinterpreted as a negative
number.
When discarding a packet for being overly long, restart the state
machine instead of attempting to drop a preposterous amount of data.
It is unlikely in that case that preposterous amount of data would
ever arrive, so rather attempt to re-align the reading algorithm
(supposedly resulting in a timeout and retransmit).
git-svn-id: svn://svn.savannah.nongnu.org/avrdude/trunk/avrdude@472 81a1dc3b-b13d-400b-aceb-764788c761c2
one byte at a time. Also mention the bug tracker interface on
savannah.
git-svn-id: svn://svn.savannah.nongnu.org/avrdude/trunk/avrdude@471 81a1dc3b-b13d-400b-aceb-764788c761c2
clash with the Win32 API names for baud rates.
Reported by: Eric
git-svn-id: svn://svn.savannah.nongnu.org/avrdude/trunk/avrdude@462 81a1dc3b-b13d-400b-aceb-764788c761c2
. Implement the new EECRAddress field in the device descriptor that is
required by the 4.x firmware; make an uneducated guess about what
firmware requires what length of device descriptor -- perhaps Atmel
can be convinced to publish an official matrix for that.
. Specify EECR in the config file where required. Obviously, only
locations that differ from the 0x3c default are mentioned in the
XML files, so by now, this only affects the AT90CAN128 for us.
. After clarification with Atmel, EnablePageProgramming should really
default to 1, and only cleared if specified by an XML parameter. So
far, only the XML files for the ATmega256x and ATmega406 do specify
it at all, and they specify a 1, too.
. Drop the entire OCDEN fuse heuristic. If OCDEN is unprogrammed at
startup, issue a warning that single-byte EEPROM updates won't be
possible. Leave it to the user to program the fuse if desired.
That way, we won't run into any issue of prematurely wearing out the
hfuse EEPROM cell. Interestingly enough, this also solved the
problem of the target not restarting from scratch upon sign-off.
git-svn-id: svn://svn.savannah.nongnu.org/avrdude/trunk/avrdude@461 81a1dc3b-b13d-400b-aceb-764788c761c2
yesterday's changes (i.e. before rev. 1.10 of ser_posix.c), that is,
exit(1) in case of a timeout. Previously, the upper layers didn't see
the timeout at all.
Quite possible that some of these drivers could handle a timeout more
intelligently though. At least for the rather sophisticated STK500v2
protocol, I think it should be possible to retry the request.
git-svn-id: svn://svn.savannah.nongnu.org/avrdude/trunk/avrdude@460 81a1dc3b-b13d-400b-aceb-764788c761c2
performance.
Note that many consumers still read one byte at a time though.
This patch has once been submitted to me by Bernd Walter
<ticso@cicely.de>, minor tweak by me (mainly to get it running under
Linux, too).
git-svn-id: svn://svn.savannah.nongnu.org/avrdude/trunk/avrdude@459 81a1dc3b-b13d-400b-aceb-764788c761c2
Brian S. Dean
Joerg Wunsch