Alpha to I64 migration

[RDB_Forever]

We are moving from an Alpha platform to I64.   Whenever we try linking our FORTRAN programs, we are seeing this error,

> LINK PL_ERROR_PRT, T_OBJLIB/LIB

%ILINK-E-INVOVRINI, incompatible multiple initializations for overlaid section

section: RDB$DBHANDLE

module: RDB$PL_INIT_INW_CNTL

file: DM$SWCTL:[BPRF.DM_TEST]OBJ_CODE.OLB;1

module: RDB$PL_FETCH_INW_CONV

file: DM$SWCTL:[BPRF.DM_TEST]OBJ_CODE.OLB;1

%ILINK-W-NUDFSYMS, 95 undefined symbols:

%ILINK-I-UDFSYM, SQL$PRC10_9U69EKJV74C10VT20480

%ILINK-I-UDFSYM, SQL$PRC11_AU69EKJV74C10VT20480

%ILINK-I-UDFSYM, SQL$PRC12_BU69EKJV74C10VT20480

[And the list continues on for 95 entries]

Any ideas what this means and how to fix it?  Everything worked fine on the Alpha, and we were hoping to just migrate straight over to the I64 architecture.

[Hoff]

Search for the word incompatible in the following for resources for information the triggers for and resolutions for this error, which usually involves sorting out your psect declarations.   The following includes documentation on this error, as well as some other issues and information around porting OpenVMS apps and OpenVMS Rdb-using applications to OpenVMS I64:

http://de.openvms.org/TUD2012/customer_success_stories.pdf

http://download.oracle.com/otndocs/products/rdb/pdf/forums_2006/rdbtf06rs_14_rdb721.pdf

http://h71000.www7.hp.com/doc/83final/4548/4548pro_010.html

http://www.oracle.com/technetwork/documentation/rdb7200-rn-134690.pdf

The forum software dropped the previous reply to this question, and then the password change blew up, which dropped everything on the floor with the usual cascade of failures...   This is the reply, re-entered.    What blows up this time?

[abrsvc]

I ran into similar types of error messages with another Fortran based application.  In my case, COMMON definitions were not consistant in their naming and size which created all kinds of problems.   Also, default alignment issues can cause similar errors.

Look at the module for COMMON definitions and verify that they are actually constructed properly.  Earlier versions of the compiler (Pre F90) worked in spite of "problems".

Dan

[Steven Schweda]

> [...] COMMON definitions were not consistant in their naming and size
> which created all kinds of problems. [...]

   It's been at least twenty years since I did much with Fortran, but
isn't that what the INCLUDE statement is for?

[abrsvc]

Perhaps, but much code exists that doesn't utilize that option (include statements).

Also, I neglected to mention that the issue is with NAMED common sections along with BLOCK DATA modules.

Dan