On Thu, 12 Jul 2018 09:40:50 +1000, Andrew Rowley wrote: > >On a unix system, you can open a file for writing and another process >can delete it while you have it open and create a new file with the same >name. The file you are writing disappears when you close it. > That's a sort of LUW isolation, a facility that has come relatively recently to z/OS in PDSE members. For better or for worse, however you view it. Is it any worse than if the processes operated sequentially; no overlap. Second guy wins. Conscientious second guy will create the file with O_EXCL. If you don't trust the second guy, lock him out with RACF profile or file permissions.
If I want to be very nondisruptive I write to a temp name and rename at the end. This guarantees that no other job sees the update in transit. UNIX rename() is preemptive: it quietly replaces any older file with the same name; and atomic: no process will perceive an instant when the file appears not to exist. I wish I could get the same behavior from IDCAMS RENAME. > ... This sort >of thing is considered bad on z/OS. Sure, that's a function of z/OS >enqueues etc., but JCL is the (relatively) easy to use interface to >allocation. On Thu, 12 Jul 2018 09:49:41 +1000, Andrew Rowley wrote: > >This works OK except when it doesn't. I recently encountered a problem >where the command implicitly sourced /etc/profile. Which meant that >things broke when /etc/profile changed the value of the environment >variable. > That sounds like a "Don't do that!" Don't do that. -- gil ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
