On Wednesday, 10 June 2015 at 18:13:53 UTC, Dave wrote:
Another backwards annotation is nothrow. I don't really care if
something doesn't throw, I care when it throws, because then I
have to do
something (or my program may crash unexpectedly).
I recently debugged such "no crash" bug: the code decided that
the program shouldn't crash and caught exception and silenced it,
the program indeed didn't crash, but misbehaved. It was a
critical bug, which blew into the face of the customer, there was
nothing in the log, we had to connect to the customer's database
and debugged with catching first chance exceptions. What we
should do if we had no access to the customer's database? If the
code wouldn't catch the exception, the application would crash
and we would have an entry in the log and debugged it quickly.
That's how nothrow works in practice.
