bug in my code?

[Terry Reedy]

"Duncan Smith" <buzzard at urubu.freeserve.co.uk> wrote in message
news:bevl84$880$1 at news7.svr.pol.co.uk...
> Greetings.  I am struggling to track down what I presume is a bug in
my
> code.  The trouble is that it only raises its head occasionally.  I
have
> managed to prune the code to the following:
>
> shutTest.py
> --------------------------------------------------------
> import cPickle
>
> def test2(iterations):
>     for i in range(iterations):
>         try:
>             f = file('C:\\Python22\\(2, 2, 2, 3, 4).shu', 'r')
>             try:
>                 dependencies, nodes = cPickle.load(f)
>             finally:
>                 f.close()
>         except IOError:
>             pass
> ----------------------------------------------------------------
>
> Calling the function with a suitable argument (say, 100) doesn't
usually
> pose a problem.  The file is opened OK and the data unpickled.  But
> occasionally (or usually if I call the function with 1000 as
argument) I get
> the all too common, on Windows,
>
> 'The instruction at {some memory address} referenced memory at {some
other
> (and on one occasion the same) memory address}.  The memory could
not be
> "read".'
...
This sounds like maybe a hardware problem.

Is the unpickling necessary to get the error? -- or some discrepancy?

If you simply reread the file over and over, do you always get
*exactly* the same byte string?  Test this by reading once before the
loop and comparing subsequent reads to first.  What I asking is
whether you might have flacky or overheated disk drive giving
intermittant errors more often than once a terabyte.

TJR