Don't want to beat this thing to death but FWIW I meant "ABEND" in the sense I hear it usually used: to abnormally end, to blow up, to go kaput. When someone says "payroll ABENDed last night" they typically in my experience don't mean it took an ESTAE exit and recovered transparently. They mean it went ka-boom. That is the sense in which I meant "ABEND not acceptable to management." I did not mean that they would care one way or the other whether the program used ESTAE and a recovery routine internally and fairly transparently.
Charles -----Original Message----- From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On Behalf Of Peter Relson Sent: Sunday, April 5, 2020 7:01 AM To: IBM-MAIN@LISTSERV.UA.EDU Subject: Re: ESPIE question (does ESPIE "cover" ATTACH'd sub-tasks) <snip> Let's posit that ABENDing on the first such condition is not acceptable to user management. </snip> That is a bad thing to posit. That is an insane design in the absence of other information, unless by ABEND you also include "and the task terminates" (and I am assuming you are including "program check" as part of "ABEND"). And any type of recovery can be used to avoid task termination due to abend (ignoring CALLRTM with RETRY=NO or things like cancel). As has been said, it comes down to performance and diagnosis. Your needs might vary. Program-check with resume (for ESPIE) or with retry without recording (for ESTAE) are pretty much invisible, aside from cycles consumed and system trace entries, so most would not care if they happened because they would do no damage aside from consuming cycles. I think that (E)SPIE was created for compiler-based exception handlers (e.g., ON conditions in PL/I). (E)SPIE did accept program interrupt codes for x'10' (segment fault) and x'11' (page fault). Perhaps Shmuel knows/remembers of some rationale, because we don't -- we cannot think of any rational scenario in a machine that supports paging where a compiler could generate code to do something useful based on getting a segment fault or page fault when that would bypass the system's processing to page in valid virtual storage that was simply paged out. I'm doubting that the compiler would be generating code to do some sort of VSMLOC/VSMLIST to ascertain if the area is truly valid. Peter Relson z/OS Core Technology Design ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
