On Mon, Feb 28, 2005 at 02:26:55PM -0800, Brian Lynch wrote:
[snipped]
Not much to go by in those traces, really.
I noticed in the first report you sent, that it seems as if hobbitd crashes about 5 minutes after it is started. This would coincide with the time when it sends in a status report about itself. Can you verify if this is correct - what is the interval between the "Task hobbitd started" messages ?
If it's 15 minutes, then it points more in the direction of the checkpoint code, and the errors you showed initially "Too few fields in record - found 1, expected 17" also point in that direction.
You could also try enabling the memory-debug code: In lib/memory.h, change the line
#undef MEMORY_DEBUG to #define MEMORY_DEBUG 1
Then do a "make allclean; make; make install" and restart hobbit. It will still crash, but hopefully a little earlier before it smashes the stack - so gdb will hopefully give some more info in the backtrace.
Regards, Henrik