Kenneth,

I know that there are some issues relative to mixing object code compiled at
5.5 with system or user libraries compiled on or supplied with 6.5 - and I
think I recall some of those were date issues.  All this is available on the
HP support site (if you can get there from here!).  I believe it is
generally safe to recompile all user modules at your current OS release
level, unless there are changes to the compiler that will not work for you.

-dtd

> -----Original Message-----
> From: Sletten Kenneth W KPWA [mailto:[log in to unmask]]
> Sent: Friday, January 05, 2001 8:40 PM
> To: [log in to unmask]
> Subject: 5.5 to 6.5EX/PP2: Date intrinsics & Transact problem...
>
>
> FYI All Transact users:
>
> Last week I did update from MPE 5.5 PP6 to 6.5 EX/PP2.
> There was a definite change in behavior for at least two of
> the HP date intrinsics, when using PROC calls from NM
> Transact programs (these programs were compiled long
> before the update;  and had been running fine without any
> change since compile).  The two date intrinsics of interest
> are:   HPDATEVALIDATE and HPDATEOFFSET.  In going
> from 5.5 to 6.5 EX/PP2 we went from NM Transact A.06.02
> to  A.07.01....
>
>
> QUICK SUMMARY:     Looks like this is a problem only for
> Transact:  HP RC confirmed our failure mode, but when they
> did same intrinsics with COBOL in same way that failed for
> Transact they worked fine...  hmmm....:  Maybe it's more
> accurate to say that as of right now NM Transact is the only
> language we know of that has this "new behavior" problem
> with HPDATEVALIDATE and HPDATEOFFSET...
>
>
> A LITTLE MORE DETAIL:
>
> HPDATEVALIDATE:   Looks like this is a problem ONLY if
> dealing with TWO-DIGIT target dates.  In that case what was
> an optional parameter on 5.5 appears to have "become
> required" on 6.5.  At least for Transact, leaving it out always
> fails.  Our site worked around this problem by including and
> initializing all optional parameters.
>
> HPDATEOFFSET:   This appears to be a problem only if you
> pass the intrinsic a ZERO offset.  Prior to 6.5 if you fed the
> intrinsic a zero offset from Transact it would leave the target
> date alone;  no change.  *BUT*:  After update to 6.5, same
> zero offset BLANKS OUT the target date;  or changes it to
> something that is unreadable / non-valid (bad, bad).  Worked
> around this one by checking for zero offset first, and not
> calling the intrinsic at all if  offset = zero.
>
>
> We were fortunate that our use of these two date intrinsics
> was fairly well localized;  enough so that my team leader was
> able to fix all occurrences and recompile all affected code
> in one day....   In any case, if you are calling these HP date
> intrinsics from NM Transact, before updating from 5.5 to 6.5
> EX/PP2 might want to check for missing optional parameters
> and zero offsets;  unless you get and install a patch to fix
> the problem along with the update.....
>
> The RC has filed CR  JAGad44611  for us against Transact,
> as a serious defect.  No info yet on when this might be fixed.
>
> Ken Sletten
>