README -- file_output
 
 2/16/2004  D. Lawrence
 
 The files in this directory implement the DAQ's file output
 naming scheme. This is a pretty convoluted system so I'll try
 and describe it here:
 
 In order to have CODA automatically split the run into files
 of a specific size, several things must be done:
 
 1. The CODA configuration in the msql database must have
    {@split_file} {CODA} for the "code" field of the coda_0
	 line in the main table for the configuration. Note that
	 this means EVERY configuration that wishes to split files
	 must have this.

 2. There must exist an entry in the "*_option" table for the
    configuration with name equal to "dataFile and value equal
	 to "@split_file". Again, this must be done for EVERY
	 configuration.

 3. There must exist an entry in the "*_option" table for the
    configuration with name equal to "SPLITMB and value equal
	 to the number of megabytes to split on. Typically, this
	 is just set to 2000.

 4. The file "split_file.tcl" must be passed to the event recorder
    when it is started. Use the -f option for this. An example
	 is given in split_file.tcl.

 
 The output_disk.pl script is run from split_file.tcl to check that
 there is room on the raid disk before we create a new file there.
 If not, then the link /home/primex/raid is switched to point to
 an empty disk. (This feature is currently disabled in the testlab
 setup since we don't have multiple raid disks and /home/primex/raid
 is a real directory.)
 
 
 DATABASE INSERTS/UPDATES
 
 An entry for each file is inserted into the Run_Files table of
 the primex_online MySQL database from split_file.tcl. New inserts
 are done using NULL for the "Number_of_Events". A special program,
 "count_events" was written to quickly find the number of events
 in a file by looking only at the front and back. (Source code is
 in /home/primex/DAQ/src/count_events.)
 
 There are two places where the number of events are updated for
 entries in the Run_Files datbase. The first is in split_file.tcl
 itself. This is so the entries will be updated as the files are
 closed and the Time field will be at least semi-accurate. It also
 will help make sure at least some files for a given run will have
 this information in the DB should the DAQ crash.
 
 The second place where "Number_of_Events" is updated is in the
 end.pl script. This is needed because split_file.tcl is not called
 when the run is ended and the last file is closed.
 
 WARNING
 
 As stated at the top of this document, this system is rather
 convoluted with scripts calling scripts and executed shell
 commands, any one of which could fail. In particular, Tcl is
 really bad about choking and dying if every little thing is
 not perfectly in place. That will make this a hard system to
 debug when an error occurs (which is why I wrote this).