case@860800.UUCP (Bill Case) (06/17/88)
Has anyone found any reasons to relink or recompile ordinary application programs under VMS 5.0? I have the 5.0 Release Notes and New Features manuals which documents potential problems with writable global sections, but I was wondering if there are any advantages to going to effort of relinking and (gasp!) recompiling. It seems like you'd be running a bigger risk of making your VMS 5.0 images unusable on VMS 4.x systems. Bill "Born to be Mild" Case For every problem there is one solution which is simple, neat, and wrong. H.L. Mencken ...uunet!ingr!b11!case!case (UUCP) ingr!b11!case!case@uunet.uu.net (ARPANET)
cole@sas.UUCP (Tom Cole) (06/20/88)
In article <65@860800.UUCP>, case@860800.UUCP (Bill Case) writes: > Has anyone found any reasons to relink or recompile > ordinary application programs under VMS 5.0? I have the 5.0 > Release Notes and New Features manuals which documents > potential problems with writable global sections, but I > was wondering if there are any advantages to going to > effort of relinking and (gasp!) recompiling. It seems > like you'd be running a bigger risk of making your VMS 5.0 > images unusable on VMS 4.x systems. > > Bill "Born to be Mild" Case Well, we happen to develop a large (read "monolithic") application, consisting of almost 5 million lines of code. It is contained in about 300 separate images, which we dynamically load into memory using the unsupported and ill-documented SYS$IMGACT service. We even unload the parts we are done with during execution. Anyway, the point is that we have a really big software package that even does a few non-standard things. We are able to run a version of this that was linked under VMS 4.2 on a VMS 5.0 system with no relink or other change, and it passes our initial validation suites. Digital (who has it now) tells us that we might gain a small performance improvement by relinking, but of course then the stuff wouldn't work in pre-5.0 systems. But other than the minor and nebulous performance increase (in the area of initial page faulting, I believe) the sucker seems to work fine with no change. The point is, I really believe them (this time) when they say that you don't need to do anything special/wierd to use 5.0... Tom Cole SAS Institute {anywhere}mcnc|rti|sas|cole These opinions are *definitely* mine and not the official position of SAS Institute (yet). As soon as they agree with me, I'll let you know.
jayz@cullsj.UUCP (Jay Zorzy) (06/21/88)
From article <65@860800.UUCP>, by case@860800.UUCP (Bill Case): > Has anyone found any reasons to relink or recompile > ordinary application programs under VMS 5.0? I know of no need to recompile code for 5.0. You must relink any privileged code, that is any code linked with the VMS system symbol table (SYS$SYSTEM:SYS.STB). We also found we had to relink our code that was linked with the TPU callable routines in SYS$LIBRARY:TPUSHR.EXE. Any images linked with the V4.x version of TPUSHR get the ident mismatch error with the V5.0 version of TPUSHR. This is because DEC decreased the gsmatch major id in TPUSHR from 5 to 2. I guess this more clearly identifies the new TPUSHR as version 2.0 of TPU, but don't ask me why DEC *down*-rev'd a run-time library. Jay Zorzy Cullinet Software