On Thursday, 07/16/2009 at 05:26 EDT, Colin Allinson <[email protected]> wrote: > However, Rob is right, that there are probably an significant number of > variations of what people may want. This might make the architecture of such > exit calls difficult. > > My suggestion would be to reverse this and externalise all the IBM code from > the SYSPROF EXEC in such a way that it is not fiddled with.
I'm thinking: 1. Codify the things SYSPROF does and formalize them. IBM._Exit_Callback = "BEFORE AFTER LAST" IBM._Exit_Name = "SPROFXIT" IBM._SYSPROF_Debug_Level = 0 IBM._CMSBATCH_Disk_A = "195" IBM._SCOMDIR = "SCOMDIR NAMES *" IBM._UCOMDIR = "UCOMDIR NAMES *" IBM._Segments = "PIPES RXSOCKET VMLIB" /* Will be loaded SYSTEM */ IBM._CSL = "VMLIB" /* Will be loaded with SYSTEM GROUP VMLIB */ IBM._FINAL_GLOBALV_INIT = 1 /* Examples of what you might see */ IBM._CMS_Fingerprint = "z/VM V5.4.0 2009-06-03 14:32" IBM._AutoReIPL = "DWAITPSW 000A0000 80F3F752" IBM._IPL_Command_PARM = "AUTOCR INSTSEG NADA SAVESYS CMS FILEPOOL" IBM._IPL_Error_Summary = "INSTSEG-LOADFAIL=NADA FILEPOOL-ID-MISSING NO-SSTAT NO-YSTAT SAVESYS-CONFLICT" IBM._Batch = 0 IBM._AutoCR = 1 IBM._DISK_192_State = "AUTOFORMATTED-AND-ACCESSED-D" : <you get the idea> 1. Invoke the exit "FIRST_CALL" 2. Invoke the exit "BEFORE_VMREAD" 3. Invoke the exit "AFTER_ACCESS" 4. Invoke the exit "LAST_CALL" The exit can examine or alter the variables as desired. (One can dream, right? I know - not on Thursdays. sigh.) Alan Altmark z/VM Development IBM Endicott
