On Thu, May 14, 2009 at 1:16 PM, Adam Badura <abadura@o2.pl> wrote:
Still there are reasons. In cases someone wants extra care he/she cannot use Boost.Exception data injection because this always adds risk of double failure - lack of memory and std::bad_alloc
This isn't double failure -- your program correctly attempts to do something (throw an exception, in this case) and fails, which results in bad_alloc exception. There is no difference between this situation and a bad_alloc you could get if you attempt to copy a string (for example.) Also, some compilers allocate exceptions on the heap. I don't see 15.1 specifying behavior in case this allocation fails, so clearly an attempt to throw an exception may fail. In this case, whether you get a std::bad_alloc or abort() won't make much difference, either way your program isn't going to throw the exception you wanted to throw (any expert care to comment on this issue?)
So to sum up IMHO if the implementation already uses this mechanism and all the required work is to make it public and document it then I would do it even if I didn't see practical use at the moment
Maybe I wasn't clear: I am not against making this change in Boost Exception, I am against doing it without understanding the use cases that require the change. Emil Dotchevski Reverge Studios, Inc. http://www.revergestudios.com/reblog/index.php?n=ReCode