64-bit OpenSSL 3.0.12 (et al.) from openssl.org v. VMS

Talk about commercial or opensource products that already exist for OpenVMS or may be available in the future.
Post Reply

Topic author
sms
Master
Posts: 349
Joined: Fri Aug 21, 2020 5:18 pm
Reputation: 0
Status: Offline

64-bit OpenSSL 3.0.12 (et al.) from openssl.org v. VMS

Post by sms » Mon Dec 04, 2023 12:16 am

Code: Select all

   Recently, I was looking at building my usual (obsolete?) GnuPG kit on
x86_64, using a recent home-built cURL kit (for the GnuPG keyserver
programs), and a home-built OpenSSL kit for cURL (as well as the
VSI-supplied cURL/[Open]SSL stuff).

   For completeness, I also tried to build cURL with 64-bit pointers,
and ran into trouble with the corresponding home-built OpenSSL kit. 
(Now two projects away from what I was trying to do.)

   When 64-bit pointers are requested on VMS, the OpenSSL file specs for
the object libraries and shared images get a "64" added to them.  (Just
like the VSI scheme, except for the many differences.)  For example:

      OSSL$libcrypto03_shr.EXE  ->  OSSL$libcrypto03_shr64.EXE
      OSSL$libssl03_shr.EXE;1   ->  OSSL$libssl03_shr64.EXE

   That seems to much work.  Sadly, the file specs and logical names in
the OSSL$INSTROOT:[SYS$STARTUP] DCL scripts do not get the
desired/expected "64" modifiers.  Apparently, this has been true for
some time, back to at least 3.0.7.

   Note that this bug affects _all_ VMS 64-bit-pointer builds, not only
x86_64.

   On a different topic, the list of configurations recognized by
config.com didn't include "vms-x86_64-p32" or "vms-x86_64-p64", only
plain-old "vms-x86_64", so specifying "-32" or "-64" to config.com was
fatal on x86_64.  (I just copied the IA64 stuff for x86_64, so if
fancier compiler options would be helpful on x86_64, you're on your
own.)

   I've logged a complaint at:
      https://github.com/openssl/openssl/issues/22899

   No response yet, but, in the mean time, if anyone's interested in a
small pile of loose files with suggested changes:

      http://antinode.info/ftp/openssl/3_0_12/

   I could be wrong, but these changes seem to me to be improvements,
and I haven't seen them elsewhere.  No bets on whether any of them
(other than the inapplicable DCL scripts) might have any value to what
VSI does with OpenSSL.

Post Reply