Nikos Mavrogiannopoulos nmav@gnutls.org writes:

In some systems the elliptic curves part of openssl is disabled. This patch disables the openssl examples/tests if it does not exist (I had issue with the EC tests).

Actual patch missing?

Also in some issue I replaced an #if with and #ifdef, because the rule didn't work. -#if WITH_OPENSSL +#ifdef WITH_OPENSSL

Hmm. I'd like to understand what the problem is.

In general I noticed a lot of uses of "#if" that actually mean ifdef. Are these intentional?

I think I have tried to use plain #if consistently. I don't remember the reason, maybe it was recommended in the autoconf manual or the GNU coding standards.

I'm not really sure what #if does if followed by an undefined expression.

When evaluation the argument to #if, I think the preprocessor replaces all unknown symbols by 0. So #if vs #ifdef should matter only if the symbol is #defined, but the definition is zero or empty or strange in some other way.

In config.h, all defined symbols should be defined to have the value 1, which should make a plain #if work fine. Maybe there's something wrong with how WITH_OPENSSL is handled?

Regards, /Niels