As I recall the x'81' in the undefined area was a mod by SLAC to the Linkage Editor. I remember it well. Mainly because IBM had a bunch of DS instructions in the MVS CVT. When I did a sysgen one time the system died because the "default" vol of 0x00 in the DS areas were now x'81'. The NIP logic apparently had a compare against 0 to see if some field was initialized and bypassed further initialization if non-zero. When the slack bytes were x'81', MVS died during NIP.
Long ago. This is my best recollection. On Thu, 2011-10-27 at 12:20 -0700, glen herrmannsfeldt wrote: > There has been an ongoing discussion on comp.lang.fortran on the > initialization of variables, including questions about OS/360 and > its compilers. > > As I understand it, Fortran variables, and DS in assembler, > generate holes in the object program (no TXT record for that > position), and are filled in either by the linkage editor or > program fetch. > > I believe that early OS/360 versions left whatever happened to > be there, either in the linkage editor buffer or, for program > fetch, in that memory location. At some time later, possibly > for security reasons, this was changed to initialize to zero, > or maybe something else (such as X'81'). > > The actual question is: when was this change made and which > programs (compilers, linkage editors, program fetch) were > changed? > > -- glen > > ---------------------------------------------------------------------- > 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 -- John McKown Maranatha! <>< ---------------------------------------------------------------------- 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
