[Numpy-discussion] mmap bug

Ralf Gommers ralf.gommers@googlemail....
Thu Jun 17 08:41:29 CDT 2010


On Fri, Jun 11, 2010 at 8:00 AM, Geoffrey Irving <irving@naml.us> wrote:

> Hello,
>
> If I create an mmap'ed array, and then generate another array
> referencing only its base, destruction of the original mmap'ed array
> closes the mmap.  The second array is then prone to segfaults.
>
> I think the best fix is to put the responsibility for flushing the
> mmap onto the actual mmap object (and therefore the base object)
> directly, so that the numpy memmap object has no cleanup
> responsibilities.  A patch to core/memmap.py follows.  It would be
> nicer to put the mmap_flush class declaration somewhere outside the
> function, but that conflicts with the local "import mmap."
>
>
Reported in http://projects.scipy.org/numpy/ticket/1513 so it doesn't get
lost.

Ralf
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.scipy.org/pipermail/numpy-discussion/attachments/20100617/641db1b4/attachment.html 


More information about the NumPy-Discussion mailing list