[openssl-dev] OpenSSL version 1.1.0 pre release 1 published

Richard Levitte levitte at openssl.org
Tue Dec 22 14:49:01 UTC 2015


In message <20151222153437.192023un5jh69v25 at www.polarhome.com> on Tue, 22 Dec 2015 15:34:37 +0100, Zoltan Arpadffy <zoli at polarhome.com> said:

zoli> Hi,
zoli> 
zoli> > zoli> May I ask you, if the new build will cover the long names issue
zoli> > (
zoli> > zoli> symhacks.h ) too?
zoli> >
zoli> > It's been pointed out to me by the vms-ports folks that it should be
zoli> > possible to solve using the compiler's "#pragma names shortened"
zoli> > rather than maintaining symhacks...  Then, it's just a matter of doing
zoli> > the same thing manually when producing the SYMBOL_VECTOR for a
zoli> > shareable image.  I'm also thinking that "#pragma names as_is" should
zoli> > be norm and that we could produce upper case aliases in SYMBOL_VECTOR
zoli> > (you know how that's done, right?).  Does that sound like a way
zoli> > forward to you?
zoli> 
zoli> It is not my decision, but I don't like either of these approaches.
zoli> 
zoli> The code itself needs to be written that it is as much as possible
zoli> portable.

In the end, this is a LINKER vs rest of the world issue.  In this day
and age, the 31 char limitation is a bit aged, and LINKER should
really be updated to allow longer symbol names.  If nothing else, C++
symbol munging would require it.  I certainly hope VSI receives the
messages and starts dealing with it.

Either way, we're dealing with C, with case sensitive symbols and
symbols longer than 31 chars.  The names pragmas are helpful
portability tools, the smoothest at this point is to put them to good
use.

zoli> It is not impossible to maintain a code base that uses up to 32 char
zoli> long function names - without losing the readability of the code.
zoli> I agree that it requires some extra focus from the developers side -
zoli> but coding a security software needs that (and even more) focus
zoli> anyway.

That can certainly be argued...  Again, that's more of a LINKER issue
than anything else.

Cheers,
Richard

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


More information about the openssl-dev mailing list