A lot of warnings on MSVC due to protected destructor being non-virtual

[Raffi Enficiaud]

See pull-request ​https://github.com/boostorg/test/pull/41

The fact that the destructor is non virtual looks like a bug to me. This was introduced in 2005, rev ef3fe4a826ebc31afbbb4c781dd3d9e504f360f1, with a commit log message "dll support introduced". It sounds like it has connection with the visibility of the destructors across dll/so boundaries, maybe for some broken compiler.

I am in favor of removing this workaround.

[Jürgen Hunold]

Full log "virtual.zip" attached. Commandline:

..\..\..\b2 -a cxxflags=-we4265 -j3 2>&1 | "C:\Program Files (x86)\git\bin\tee.exe" virtual.log

See ​https://msdn.microsoft.com/en-us/library/wzxffy8c.aspx for the official description. We normally make this a level 3 warning, I've changed this to an error to get clearer test result.

[Raffi Enficiaud]

Thanks. So I'm not addressing the trillion of warnings from error_info_container in boost.exception.

The warning for collection is indeed a true one, good catch! I am adding a virtual destructor to class dataset, which should solve most of the issues in your log.

However, this has nothing to do with the pull request in fact (and Gennadiy was right). The macro that gets activated with your change does not solve the problem I can see in this log.

[Jürgen Hunold]

virtual destructor error log

[Jürgen Hunold]

Sorry, I had the patch still active when doing the first run. Uploaded the correct run.

[Raffi Enficiaud]

A quick look at this tells me that most of the warnings can be safely ignored: in fact most of the instances are not meant to be used on the heap, but rather on the stack.

I will have a deeper look tonight/tomorrow.

[Raffi Enficiaud]

Addressed by rev 13ca3e07f071ee36a0c5be5c6a0b58be7ab4637c