On Fri, 24 Mar 2017 14:26:14 -0500, Joel M Ivey <[email protected]> wrote:
. . .
>+972 4140 0004
>+976 1904
>+978 4770 C986 <-----
>+97C 5802 0000
>+980 41F0 0004
>+984 07FE
>
I don't have the luxury of your assembly listing, but the ESTAE in the source
code from cbttape.org looks like this:
RECOVERY DS 0H 08440000
USING *,R15 SET UP ADDRESSABILITY 08450000
LA R4,4 PUT 4 IN REGISTER FOR COMPARE 08460000
CR R0,R4 IS SDWA PRESENT? 08470000
BNE HAVESDWA YES, BR TO PROCESS WITH SDWA 08480000
L R0,0(R2) LOAD RETRY ADDRESS FROM PARAM LIST 08490000
LA R15,4 SET RETCODE TO RETRY ADDR IN R0 08500000
BR R14 RETURN TO PROCESSOR WITH RETRY ADDR 08510000
HAVESDWA DS 0H ENTER HERE IF SDWA PRESENT
That code mirrors what support found, *except* it would appear that it is
assembled from code which has the " USING *,R15 " statement missing (in the
object code, the BNE instruction is using R12, which is a mainline base
register, not R15 as it should).
That makes me think that for some reason you are perhaps *not* executing the
version of IEFACTRT that you think you are, and that installing a new one, from
SAMPLIB or elsewhere, is not going to help you until you figure out exactly why
that is happening. For a start, check your listing. You might have received a
warning about an overlapping USING, but I think the correct code should still
have been generated in this instance for the BNE instruction.
Over the years I saw problems like this many times when recovery (or retry)
routines made unwarranted assumptions about the content of certain registers,
particularly base registers, and the warning about the possibly dire
consequences is justified. One such case is still firmly etched in my memory,
well over ten years later - the wild branch there happened to take it part way
into an instruction, which unfortunately was executed as a perfectly valid LCTL
instruction. The results of loading rubbish into half of the control registers
is not pretty.
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
