On Fri, 29 Jul 2005 13:56:21 -0700, Edward E. Jaffe wrote: >How about the wild branch diagnosis improvement? (Does anyone on this >list still write code?) > >z9 109 remembers the last successful branch instruction address and the >operating system displays that address in dumps. I would guess that at >least 90% of S0C1 abends are caused by so-called "wild branches" ... a >large percentage of which end up at location zero. Given that, a typical >debugging scenario is to look for zeros in R15. If so, assume the return >address in R14 is the last known address. But what if it isn't? What if >all of the registers are "clobbered"? There are many different possible >wild branch outcomes and some can be *extremely* difficult to diagnose >(BTDTGTS). > >My reaction? It's about time! Thank you, IBM!
My reaction is similar to yours, Ed. I had been wondering about the need for some kind of "come from" display when IBM brought branch relative long instructions to the architecture. The R15/R14 game seems like it falls apart if the arrival at location 0 (or just arbitrary low storage) came about because of a branch relative (jump) or worse, a BR-long. Having the last successful branch displayed in dumps will help (as long as the "branch from on high" isn't followed by a random branch down low before the abend strikes; then all bets are off without an itrace). Frequent culprits are the access method GET/PUT routines not being filled in (due to OPEN failures that went unchecked) and those cases seem like they'll be covered by the 'wild branch' display. (Oh, joy!) -- Tom Schmidt Madison, WI ---------------------------------------------------------------------- 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
