On 24/07/2015 16:19, Daniel P. Berrange wrote: >> > Right... We need to introduce CONFIG_CRYPTO and use it in >> > tests/Makefile too. > Except it was a goal of this work that we do not have such config > parameters leak outside to callers.
CONFIG_CRYPTO would disable all of crypto/ if neither tools nor system emulators are compiled (e.g. using $(call lor) and $(call notempty)). It would be an internal detail of the Makefile. Paolo > The intent was that the APIs > are always built-in - we just conditionally disable the internal > implementation, so that code using these APIs does not need to > deal with conditionally disabling itself like this. It is > preferrable to just fix disablement of gnutls IMHO.
