[openssl-dev] OpenSSL version 1.1.0 pre release 3 v. VMS

[Richard Levitte]

In message <16021619150191_20205A47 at antinode.info> on Tue, 16 Feb 2016 19:15:01 -0600 (CST), "Steven M. Schweda" <sms at antinode.info> said:

sms> From: Richard Levitte <levitte at openssl.org>
sms> 
sms> > sms> Configuring for vms-alpha
sms> > sms> %DCL-W-MAXPARM, too many parameters - reenter command with fewer parameters
sms> > sms>  \2\
sms> > 
sms> > Yeah, that one is entirely harmless (and known).
sms> 
sms>    Ok.
sms> 
sms> > sms>    Apparently, the test for "--prefix" is case-sensitive, and DCL can't
sms> > sms> be trusted.  Quotation helps:
sms> > sms> 
sms> > sms> ALP $ @ config.com "--prefix=ALP$DKC100:[UTILITY.SOURCE.OPENSSL.1_1_0-pre3]"
sms> > sms> [...]
sms> > 
sms> > You're running with the extended parsing style, aren't you?  I always
sms> > forget that possibility ...
sms> 
sms>    Of course.  Isn't everyone?

Nope.  I'm a traditional kinda guy in that regard ;-)
(frankly, I've been away from the VMS scene for a chunk of years.  I
do recall the parsing style thing, but that was fairly new back then
and nothing I cared about too much...  I may learn to like it soon
enough, just let me catch my breath a bit first while I catch up ;-))

sms> > sms>    Installation fails, however:
sms> > sms> 
sms> > sms> ALP $ mms install
sms> > sms> 
sms> > sms> *** Installing development files
sms> > sms> CREATE/DIR ossl_installroot:[include.openssl]
sms> > sms> COPY/PROT=W:R openssl:*.h ossl_installroot:[include.openssl]
sms> > sms> %COPY-F-OPENIN, error opening openssl:*.h as input
sms> > sms> -RMS-F-FNM, error in file name
sms> > sms> %MMS-F-ABORT, For target INSTALL_DEV, CLI returned abort status: %X1067109C.
sms> > sms> ALP $ 
sms> > sms> 
sms> > sms>    I haven't yet looked into that one.
sms> > 
sms> > When looking into it, I suggest you do so with
sms> > mms/macro="NODEBUG=""""", that'll display the temporary logicals that
sms> > get created under the hood.  (fair warning, it's not the most elegant
sms> > coding)
sms> 
sms>    I'll see if I can work up some ambition.  Before it died, the
sms> installation procedure seemed to create a new/strange/bad directory for
sms> the header files:
sms> 
sms> Directory ALP$DKC100:[UTILITY.SOURCE.OPENSSL.1_1_0-pre3.include.openssl]
sms> 
sms> aes.h;1                    11  15-FEB-2016 20:33:03.39  (RWED,RWED,RE,R)
sms> [...]
sms> 
sms>    The [.openssl] subdirectory looks new (and undesirable).  (And,
sms> nothing beyond the header files got installed.)

That's perfectly intentional and correct.  The day there comes a
compiler that knows how to concatenate directories properly (the fact
that DECompHP C doesn't know how to do that is beyond me and one of my
pet peve among many with that compiler), I certainly want for anyone
to be able to say something like:

    /INCLUDE=ALP$DKC100:[UTILITY.SOURCE.OPENSSL.1_1_0-pre3.include]

or possibly:

    /INCLUDE=ALP$DKC100:[UTILITY.SOURCE.OPENSSL.1_1_0-pre3.include.]

and the compiler will know exactly what to do with something like this
without having to define a logical name:

    #include <openssl/evp.h>

The current need to have a logical name (such as OPENSSL:) for every
such inclusion directory is a formidable pain in the nether regions.

So yeah, there's meaning with that layout.

Cheers,
Richard

-- 
Richard Levitte         levitte at openssl.org
OpenSSL Project         http://www.openssl.org/~levitte/