Jim or Ann,
I'm debugging a serial log corruption that occurs during the
falcon_chill_thaw stress test. It seems that the length field for
SRLUpdateRecords entries is either missing or is overwritten when the
data length is zero.
Do you have any documenation on the physical encoding of the integer
values used in the serial log for transaction ids, savepoints and
length? If not, can you throw together an explanation?
I assume that the encoding is value-based, endian-independent and packed
to preserve space within the sector-size blocks...true?