On Sun, 15 Jun 2014 01:10:29 +0300, Binyamin Dissen wrote: >On Fri, 13 Jun 2014 11:11:56 -0400 David Cole wrote: > >:>The Binder, on the other hand, when he's told to mark a load module >:>as being Refreshable, he will also unconditionally mark it as being >:>also Reentrant and Reusable, and the Contents Supervisor's logic >:>relies upon this fact. > >Which technically is wrong, as ... > I believe that as long as Linkage Editor is supported, it's improper for CSV to rely on behavior peculiar to Binder.
Beyond that, I know of ISV cross-compilers that produce load modules (in TSO TRANSMIT format, I believe) according to the published specifications of load module format. At SHARE Denver, John E. (I forget which one, precisely) breathed a sigh of relief that Program Objects are now considered opaque objects and customers can no longer report errors arising from idiosyncrasies that IBM utilities never employ. Does the specification of REFRPROT require both REFR and RENT, or that the module have been created by Binder? How might a programmer regard REFR,NORENT as useful? Suppose a module never modifies itself (or is modified), so is properly marked REFR. But that module (and only that module) manipulates some external resource in a manner requiring serialization. The programmer takes a shortcut and marks the module REFR,NORENT and relies on CSV rather than some more customary technique such as ENQ to perform the serialization. But would this ever have worked, or is it aboriginal behavior of CSV to treat REFR as implicitly RENT? -- gil ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
