Category: General LinuxCNC Questions
Hi again,
The bug has been found in the driver. It was indeed not counting correctly resulting in trying to overflow the buffer (that caused an abort without writing beyond the buffer). I will be making a patch for it tomorrow.
Also, the mrje70a device is something of a piece with very complex modbus communication (I downloaded the comms spec). It seems to use "record reads and writes in depth" where several values are read/write from an address in a structure like fashion. These must be read/written using a single modbus transaction (when the "continuous read/write" column set to "impossible" it seems to denote that it is a record read/written in depth and not by incremental address). That is the first time I've seen such a hack. Luckily, the driver can do that (but you need to ignore the misalignment warnings from mesambccc). However, some extra pin setup is needed before you want to start writing or you will be sending half or malformed data. I'll show how to do that when I have made the patch.
All that said, the patch will also include a fix for the completely brain-dead mrje70a communication, where it returns a single signed byte (!) in a word register's LSB and zeros the MSB while leaving the user the hassle to tackle the sign extension. The patch will allow you to perform the proper sign extension to s32, s64 or float hal types by introducing a byte type.
