Re: File manipulations in Ada.Text_IO

[Robert A Duff <bobduff@shell01.TheWorld.com>]

Jeffrey Carter <spam@spam.com> writes:

> Larry Kilgallen wrote:
> > But such a function for a non-wildcarded filename is always going
> > to be susceptible to a race condition if the next step would be
> > to open the file.  The file might be deleted or created between
> > the test and using the information gained from the test.
> 
> That's true in the general case, and something I've never encountered in
> 30 years of actual SW development, including software that was intended
> to manipulate the same file from different applications running at the
> same time.

How do you know?  ;-)  I mean, if somebody writes code that contains
such a race condition, and their customer trips over it once in a while,
the programmer might never find out about it.

The server process crashes once in a while.  Nobody knows why.  So the
solution is to restart the server by hand.  (Yuck.)

Race conditions, by their very nature, are impossible to detect with any
confidence via testing.

- Bob