"bearophile" <bearophileh...@lycos.com> wrote in message
news:i2knnc$1ft...@digitalmars.com...
Walter Bright:
1. segfaults *are* exceptions.
Aren't exceptions objects?
Bye,
bearophile
Not at all -- exceptions are system-generated events that are implemented in
modern languages by the equivalent of a throw statement, and Exception
objects are used to communicate information about either system- or user-
generated throws. And Walter's point is getting lost -- 'retard'
misidentified the issue as giving up speed by going with "exceptions" to
avoid "hard crashes" such as segfaults, but of course there is no loss of
speed for catching a segfault signal -- an asynchronous event -- and
throwing an Exception object rather than calling a C signal catcher. The
speed cost doesn't come from using Exceptions, it comes from doing array
bounds checking (which, of course, results in an Exception object being
thrown if violated). Of course, in D, if you're desperate for it, you can
regain the speed by giving up the memory safety and using pointers, in which
case an array bounds violation might overwrite something -- and if the thing
overwritten is a pointer, that is likely to result in a segfault -- which is
an exception, which in D will cause an Exception object to be thrown. Of
course, in languages that have them, objects are used to communicate the
information about exceptions, but even in C there are (non standard)
libraries available that provide a catch/throw/exception-passing mechanism
using a stack of setjmp structs and a longjmp for the throw.
