Adam Megacz wrote:
> Jeffrey Altman <[EMAIL PROTECTED]> writes:
>> If you want your code to be file system, AFS client, and version
>> independent you really need to test the functionality and and
>> not the implementations.  Specifically, you must perform a test
>> that attempts to lock a byte range, spin off a second process attempt
>> to lock the same byte range, and test for the failure.  If the
>> failure doesn't occur, then you know you need to fall back on a locking
>> strategy that doesn't rely on byte range locks.
> 
> Absolutely.  In fact, I'd already implemented that.  The only problem
> is that it doesn't work on Linux since the Linux AFS client does a
> more elaborate job of faking the byte-range locks.

The AFS Linux client does not implement this functionality.

The byte range support you are seeing comes from the Linux kernel and it
is independent of AFS.  The equivalent AFS functionality has not been
written.

If your app is deadlocking, then there may be something wrong in the
app's lock hierarchy.

Jeffrey Altman


Attachment: smime.p7s
Description: S/MIME Cryptographic Signature

Reply via email to