[Pvfs2-developers] possible bug ... can I get another pair of eyes?

[Murali Vilayannur]

Hi Walt,
Good catch!
completion_list_retrieve_completed() should probably return MIN(i, limit).

We dont run into this right now because we
probably never have a situation with > 64 (MAX_NUM_OPS) outstanding
completed ops.
I can fix this in HEAD if noone has any objections..
thanks,
Murali

On Thu, 7 Sep 2006, Walter B. Ligon III wrote:

>
> Hey, I've been looking at src/client/sysint/client-state-machine.c in
> the function completion_list_retrieve_completed() on about line 150 (in
> my version, the current trunk will be a little different).  There is a
> variable *out_count that concerns me.  As it is written in my version
> this variable is set to "i" which should be the number of items in the
> completion list before the function is called (or to be more precise,
> the value of s_completion_list_index).  If you look at how that variable
> is USED - particularly in src/apps/kernel/linux/pvfs2-client-core.c on
> or about line 2412 - it is ASSUMED to return the number of items that
> were copied into the arrays.  This is true only so long as the original
> size of the arrays is LESS than the number of items in the completion list.
>
> I don't know if this was the intended behavior for PVFS_sys_testsome and
> the client-core is mis-using the variable, or if this is not what should
> be returned (and if its not, if it should be fixed in testsome, or in
> completion-list-retrieve-completed).  My guess is that
> completion-list... should return the MIN of "i" and "limit" rather than
> just "i".
>
> Can I get another pair of eyes to look at that and see if I'm missing
> something.  If this is a bug I'll happily fix it in my branch.
>
> Walt
> --
> Dr. Walter B. Ligon III
> Associate Professor
> ECE Department
> Clemson University
> _______________________________________________
> Pvfs2-developers mailing list
> Pvfs2-developers at beowulf-underground.org
> http://www.beowulf-underground.org/mailman/listinfo/pvfs2-developers
>
>