Quote string(REPLACE) cmake command fourth argument

[Manu343726]

This commit fixes potential errors when CONAN_VERSION variable
in conan_split_version() function is not set so the command
is interpreted as with 3 arguments only. This could also help if
the variable has spaces so the command is interpreted with 5 arguments.
As a rule, always quote cmake command arguments

See https://travis-ci.org/foonathan/type_safe/jobs/186587154#L756 and https://travis-ci.org/ChaiScript/ChaiScript/jobs/186450865#L558 for examples of this error.

[CLAassistant]

All committers have signed the CLA.

[codecov-io]

Current coverage is 94.51% (diff: 100%)

Merging #801 into develop will not change coverage

@@            develop       #801   diff @@
==========================================
  Files           228        228          
  Lines         16679      16679          
  Methods           0          0          
  Messages          0          0          
  Branches          0          0          
==========================================
  Hits          15764      15764          
  Misses          915        915          
  Partials          0          0          

Powered by Codecov. Last update 1a5eaec...6ff9e0f

[lasote]

So why the variable is empty?

[memsharded]

It could be empty, because conan CMake macros are assuming CMAKE_CXX_COMPILER_VERSION has a value. We are just checking it for the compiler version, assuming that C++ compiler will be always installed, but it could happen that it is not. So IMO it can be safely merged.

[memsharded]

But I have a question, from the output of https://travis-ci.org/ChaiScript/ChaiScript/jobs/186450865#L558, I don't understand the error, CMAKE_CXX_COMPILER_VERSION, should have a value for that case, right? Just want to understand what is happening, maybe there is some underlying issue we could better address.

[lasote]

@memsharded that's why I asked, I don't fully understand why the variable is empty in that case.

[Manu343726]

So why the variable is empty?

That's another question. What I'm really trying to fix is the usage of a cmake antipattern in your generated code. But you're right, maybe there's a deeper issue with that empty variable.