Hi all.
At least in OpenSSL 1.0.1e, rand.h header file includes the Windows
specific windows.h header and does so in such a way that it can cause
conflicts with the rest of OpenSSL code.
windows.h defines some symbols like X509_NAME & X509_EXTENSIONS that
conflict with OpenSSL types. ossl_typ.h has a workaround for this by
#undef-ing such symbols in case they were already registered by a
previous windows.h include. This, of course, assumes that the ossl_typ.h
header needs to be included after the windows.h header.
OpenSSL's rand.h header on the other hand includes first the
ossl_typ.h header and only then windows.h - causing symbol conflicts and
messy compiler error messages in case the windows.h header file actually
gets included here and not earlier.
Note that simply changing the order in which these two headers are
included will fix this specific issue but the whole system is still open
to the same sort of problems in case another OpenSSL header includes
ossl_typ.h and only then does the rand.h header get included.
I believe a better fix would be one of the following:
1. Add a compile-time error in case the windows.h header has not
already been included prior to including rand.h (that is the only place
in OpenSSL where that humongous header is included anyway) and have the
user be explicitly in charge of including it prior to including any
OpenSSL headers.
2. Remove the need for including the windows.h header from rand.h
completely by simply explicitly declaring those few windows specific
types used there. I believe it is just the UINT, WPARAM & LPARAM types
used in RAND_event() - all simple types, and not really worth including
the whole windows.h just for them...
My personal preference is the solution #2.
Hope this helps.
Best regards,
Jurko Gospodnetić
______________________________________________________________________
OpenSSL Project http://www.openssl.org
Development Mailing List [email protected]
Automated List Manager [email protected]
