From: Jens Maurer (Jens.Maurer_at_[hidden])
Date: 2001-04-25 13:00:32
Daryle Walker wrote:
> For the new test, the "bad" class uses its related objects before those
> objects are initialized. The "good" classes use the Ron's member helpers to
> get initialization in the right order.
I believe that tests (read: regression tests) should ensure that the
implementation works on the given platform. I'd feel more confident if
the tests would extend the current "compile" check to a check which
tells me that the compiler actually got the initialization order
right. The latter is something I currently have to do manually.
I think you can read uninitialized values of "unsigned char"
type, because they're required to have a (legal) value for every
bit pattern of their representation.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk