On Dec 21, 2010, at 01:24, Martin Trübner wrote: > >>> Oops. Of course, I should have said a +/- 4GiB relative >>> displacement. :-[ > > Considering the options for running/residency of the program- where > would that make a difference? ;-) > Could this happen with RMODE(SPLIT); code below and data above the bar? Might Content Supervision report an error in some cases?
On Dec 20, 2010, at 23:09, Robert A. Rosenberg wrote: > > > I had not realized that there were now RLD records that showed the > displacements for these type of cross CSECT offsets. That is what > would be needed (An offset into the CSECT with the LARL) so the > command could be located and parsed to compute the new offset when > the displacement from the LARL's CSECT to the Target CSECT is altered > by the Binder. > I suspect this could most simply be done as complex relocation; negative for the code CSECT, positive for the data. I hope intermediate results are kept as 64 bits for present and future OSes in which the bar is erased. I hope the Program Management APIs for this are documented for the benefit of ISV compiler writers. I wish, with less optimism, that the SYSLIN format is documented for the benefit of cross-compiler authors. I was dismayed at SHARE Denver to hear an IBM representative boast of the design liberty IBM has gained by making the program object format non-GUPI. -- gil
