[PVFS2-developers] PVFS2 Request question

[Rob Ross]

Hey Murali,

Thanks for working on this more and narrowing down the problem to the file 
side.

Is there a simpler case of the contig-memory/noncontig-file that would 
exhibit the same behavior?  You indicate that even for say 2 tuples the 
problem exhibits itself?

It would be helpful to us to get down to the simplest test program
possible, preferably one that deterministically does the wrong thing in 
the same way :).  I think it would be helpful for Walt to see the 
resulting vs. desired output as well.

Thanks!

Rob

On Thu, 15 Jul 2004, Murali Vilayannur wrote:

> Here are the results of what I have been able to test thus far.
> Attached are modified versions of the same program, one of which tests
> noncontiguity on files and the other on memory. For what it is worth, this 
> could be added to the test sub-directory or something.

[snip]

> b) the test that does contiguous memory + non-contiguous on file does 
> _not_ _work_ unless the # of tuples is set to 1. This is the test program mem_contig-test.c.
> The size of the file that is created seems to be incorrect.
> I also think that PVFS_Request_commit has a memory leak.
> I think there should be atleast a  free(*req) just before *req = region if 
> not more.

[snip]

> I think it is a file type problem, unless of course I have managed to 
> introduce newer bugs into my code :).
> The file size is not correct. If you wish me to take a look at the 
> contents of file, then I could do that as well and report back later 
> today.