Python handles globals badly.

[Mario Figueiredo]

On 12-09-2015 18:09, MRAB wrote:
> On 2015-09-12 17:29, Dennis Lee Bieber wrote:
>>     But no one had tested the algorithm with the rate of change the
>> Ariane
>> 5 could produce -- so an algorithm that was developed for, and safe with,
>> the smaller Ariane suddenly went "something's wrong -- abandon ship"
>>
>>     Nothing inherent in the language...
>>
> What would C++ have done in the same situation? Would Ariane still have
> failed? Probably...
> 

And that's exactly the point. C++, or Ada, for that matter have decades
old documented best practices and code patterns to deal with those
aspects of the language that can induce in error. Integer overflow is a
well documented problem. And relying on it, is documented as a bad idea
for several reasons, including the changes in the underlying system that
eventually led to Ariane incident.

For all that is worth, C++ issues with all sorts of overflows and
unchecked memory are documented from the very first beginning of the
language. Same with C and same with Ada own particular issues.

A safe(r) language just presents different ways of shooting one's foot.
We can discuss how much of a bad boy C++ is, but at the end of the day
programers will just keep on make mistakes and eventually on those very
areas the safer language doesn't provide a safety net.

One can argue that by offering more ways to shoot one's foot, C and C++
are more dangerous to use than other considered safer languages. But
that doesn't gel with the operative history of C or C++ that are running
mission critical systems, from stock markets to nuclear power plants.
These languages just demand a different breed of programmers and
different methods of testing.