Le 24/08/2015 21:19, Georg Baum a écrit :
I am a bit confused, maybe because my french is not good enough anymore;-) My current understanding is this:
Well, your French is not that bad actually.
- qt5 + --enable-cxx11 does not compile. I verified this and that the patch fixes the issue and does not break non-cxx11 builds, so I put it in.
Good.
- unique_ptr cannot be implemented safely without c++11, so as long as we want to support C++98, you cannot use it. I agree with Jean-Marc here that C++98 support in 2.2 should be maintained. My experience with backporting from a diverging master is that it is usually not difficult to do, but of course it will be more than git-cherrypick.
I do not know that much about auto_ptr and unique_ptr, but would it be possible to use auto_ptr as a poor man's unique_ptr? If yes, is it dangerous?
- something needs to be done about auto_ptr when compiling with C++11 to get rid of the warnings, but we don't know the solution yet.
It seems that some pragmas can help gcc >= 4.5 http://stackoverflow.com/questions/13459602/how-can-i-get-rid-of-deprecated-warnings-in-deprecated-functions-in-gcc
Did I miss anything?
Discussing whether we want to turn c++11 on automatically on platforms that support it.
One problem that we have is that the compiler has to be in c++11 mode _and_ the configure script has to be aware of it. Currently weird compilers are not handled at all.
JMarc
