The memory data was introduced for the sole purpose to provide an
nvm_data_offset for Microchip programmers. As an address offset it would
tell programmers to read/write in data space. It so turned out that this
nvm_data_offset always was 0x1000000 for PDI and UPDI parts.
This commit removes this artificial data memory and treats the data offset
in the code as what it currently is: a constant.
At the same time the commit introduces an sram memory with offset and size
for virtually all parts known to AVRDUDE.
* Harmonize WIN32 implementation of serial time-outs with Posix
This is from @mariusgreuel in Issue #1249.
https://github.com/avrdudes/avrdude/issues/1249
> The WIN32 time-out is specified as ReadTotalTimeoutConstant + buflen * ReadTotalTimeoutMultiplier, which makes it potentially much longer than the Posix one (i.e. at least double). ReadIntervalTimeout is also set to timeout, which essentially renders the interval timeout ineffective, so IMHO, we should explicitly disable that feature.
* Update NEWS
* Always print error if part and programmer doesn't have a common programming interface
even though the user uses -F. It won't exit though.
* Don't mention -F in error message if -F is already used
---------
Co-authored-by: mcuee <xiaofanc@gmail.com>
Co-authored-by: Stefan Rueger <stefan.rueger@urclocks.com>
adding code that allows read-only memories to appear written successfully
provided the data to be written is already in the memory. This is useful
so that "writing" a backup to read-only memories works
- Without generating an error
- Without actually trying to write to the part
Trying to write a read-only memory with a byte that is not there will
cause a write error. This is useful if for example one tries to write a
backup of a part to another part with different signature. If the
signature is written back first, this constitutes a "check we are writing
to the same type of part".
This is from @mariusgreuel in Issue #1249.
https://github.com/avrdudes/avrdude/issues/1249
> The WIN32 time-out is specified as ReadTotalTimeoutConstant + buflen * ReadTotalTimeoutMultiplier, which makes it potentially much longer than the Posix one (i.e. at least double). ReadIntervalTimeout is also set to timeout, which essentially renders the interval timeout ineffective, so IMHO, we should explicitly disable that feature.
Rather than finding memories by name this finds memories by the correct
type further separating memory names from their functionality.
Also fixes a few unchecked avr_locate_mem() calls that could lead to
segmentation faults.
