On Mon, 28 Jan 2008 08:54:45 -0600, McKown, John wrote: >> -----Original Message----- >> [mailto:[EMAIL PROTECTED] On Behalf Of Miklos Szigetvari >> Sent: Monday, January 28, 2008 8:41 AM >> >> I mean the program is ready and working , so no program >> modification, and it shoud happen "in flight" >> For example keep only the last part of the trace. > >Ah! My bad. In that case, I guess you would need to "pipe" the output to >another process which would actually implement the "wrap around" logic >to a file. I don't know of any such program, off hand. Does the "Batch >PIPES" product still exist? That might be easier to implement than doing >it yourself using UNIX facilities. The main problem that I see is >starting the UNIX receiver process when your program starts. Unless your >program is a UNIX program, then it is rather simple to start up the >"receiver" using shell piping. > Might yet be done allocating the trace file in JCL to a named pipe (FIFO) to be read by a chopping daemon. The FIFO could be allocated and the daemon started in a prior job step. Two likely problems:
o If the trace records are variable length containing binary information, that can not be preserved in a Unix file. o If the daemon crashes, it soon takes the application down with it. -- gil ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
