Technical notes on my code submission of 26 Jan 00 Kern Sibbald General: - Reviewed master/slave network problems reported by Marcus - Documentation Changes submitted this submission: - I moved the slave_count and slaves[] data out of the shared memory buffer thus reducing its size by about 6K. This was because of a problem reported by Marcus that the master would stall for about 90 seconds after a slave went down. This occurred because the shared memory buffer was being locked while the master contacted the slaves. Thus it prevented the master's serial port code from accessing the shared memory buffer. The solution to the problem was to remove the shared memory locking while contacting the slaves, and move the slave variables out of the shared memory buffer into global memory. This required a substantial number of changes to apcnet.c to remove the ups reference infront of all references to slaves[]. However, it should be noted that this was purely a mechanical change. There were also a very small number of similar changes in apcconfig.c and apcupsd.c as well as moving the definition of the variables. The only "real" change to the code was to remove the read_andlock_sharea() and the write_andunlock_shmarea(). - I eliminated slave_bool as it is never referenced. - I updated the document to correct some spelling errors and to incorporate some of the ideas learned from Marcus' tests. I updated the new features section to contain the ChangeLog - Corrected a minor problem in the example slave config file where the UPSTYPE was not the same as in the master config file example. I also set NOLOGIN to disable in each file as this is a more reasonable default.