fixed several memory leaks of local HD5Read buffers

[hainest]

Several local buffers used in calls to HD5Read were not deallocated before exiting the local scope. Most of these were added by commit a48a540.

[scopatz]

Thanks for noticing our shortcomings, Tim. Double thanks for the fix.

[crbates]

So in my quest to learn more c++ I recently stumbled across std::unique_ptr which is a nice way in c++11 to not have to worry about deletes. I'm not saying we should use it instead but I figured I would mention it here for future reference.

[hainest]

@scopatz It was minor stuff, but you are quite welcome. One other thing I noticed in cpp/data.cpp is the usage of empty catch blocks to "pass" the failure of library loads. I am assuming this hasn't caused any issues, but may warrant investigation.

@crbates Indeed. std::unique_ptr can now contain array pointers which std::auto_ptr could not. There is lots of great new stuff in C++11.

[scopatz]

@scopatz It was minor stuff, but you are quite welcome. One other thing I noticed in cpp/data.cpp is the usage of empty catch blocks to "pass" the failure of library loads. I am assuming this hasn't caused any issues, but may warrant investigation.

This is because these functions have model fallbacks (or should) if the data is not available. Moreover we can't guarantee that the data will be available on any given machine. Such differences should show up in the tests, though.

So in my quest to learn more c++ I recently stumbled across std::unique_ptr which is a nice way in c++11 to not have to worry about deletes. I'm not saying we should use it instead but I figured I would mention it here for future reference.

@crbates Indeed. std::unique_ptr can now contain array pointers which std::auto_ptr could not. There is lots of great new stuff in C++11.

My personal opinion about this is that as a rule we shouldn't be using C/C++ smart pointers. If you want that kind of memory management in PyNE, just use Python or Cython (with the public flag). The thing that differentiates C/C++ is that they don't force you to have smart pointers.

Also, I am not convinced that C++11 is really ready for the prime time yet (much like Python 3), sadly. There are things that I would love to get rid of first (like std::map), etc.