On Sat, 2 Jul 2016 19:25:16 -0500, Paul Gilmartin <[email protected]> wrote:
>On Sat, 2 Jul 2016 16:06:49 -0700, Ed Jaffe wrote: >> >>> ... If an address >>>> contains 7FFFFBAD then you know it's an intentionally "bad" address. >>>> >>> Hmmm. If you branch to it, do you get an addressing exception or a >>> specification exception. I'd guess the latter, but it might be model >>> dependent. >> >>It's not model dependent. You never get a specification exception on a >>branch. >> >??? From: > CICS Transaction Server > CICS Transaction Server 5.2.0 > Troubleshooting and support > Dealing with the problem > Dealing with transaction abend codes > What type of program check occurred? > ... >6 > Specification exception - incorrect format of an instruction or invalid > registers. > Possible causes are as follows: > Overlaid program > Incorrect field lengths used in packed decimal multiply and divide > instructions > Branch to an odd-numbered address, caused by an overlaid register save > area It may be pedantic, but it really isn't an exception generated by the branch itself. The various "branch" instructions generate very few exceptions. If you look at the PoPs you'll typically see them generating only Trace and Transaction Constraint exceptions. From the PoPs, it's true that if you branch to an odd address you'll get a specification exception. However, the exception happens later, after the branch has finished, when the system notices the odd address in the PSW. -- Walt ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
