https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86842
--- Comment #3 from Jonathan Wakely <redi at gcc dot gnu.org> --- (In reply to fiesh from comment #2) > > Not everything. Some Expects preconditions cannot be tested (e.g. that > > pointer points to an array of at least N objects). > > True. I suppose it's sufficient to have that memory accessed (somehow > explicitly or via other checks), allowing sanitizers to catch bugs here. > > > That sounds like a maintenance nightmare. > > Can you elaborate? There could also be a define like > I_DONT_KNOW_COMPLEXITY_CLASS that sets the exponent n of the O(N^n)? Every additional knob that users can turn increases the implementation burden and the testing required for proper coverage. So I'm strongly opposed to that part of the suggestion. Either you get all the ABI-preserving debug checks, or none.
