[Swift-devel] deep field bug

[Michael Wilde]

To close this issue down: Mihael provided the following second patch on top of the diff below:

  http://www.mcs.anl.gov/~hategan/deepfieldbug2.patch

This has been working for me, and appears to completely solve this problem.

Mihael, can you commit these two patches?

Thanks,

Mike



----- "Mihael Hategan" <hategan at mcs.anl.gov> wrote:

> Mike (and possibly others) have been experiencing a bit of a problem
> with a certain type of script.
> 
> The basic idea is an app returning a complex structure. The example
> Mike
> gave was (irrelevant stuff removed):
> 
> type filestruct {
>   file filefield;
> }
> 
> app (filestruct fs[]) touchem () { touchem; }
> 
> Where fs is mapped properly with a fixed array mapper.
> 
> The problem with that is that closedataset (which is called when
> touchem
> is done) only closes fs and fs[*], but not fs[*].filefield. For some
> reason the method used internally is closeChildren().
> 
> Now, it seems quite obvious that that's bogus and a deep close should
> be done instead when returning from an app. Furthermore, I see no
> need
> for closeChildren() to ever be called on its own (other than perhaps
> by
> a deep close). So I'm sending this email to see if anybody (Ben that
> is)
> is of a different opinion.
> 
> In the mean-time, for those who want to test a possible solution (and
> please run the whole test suite in the process):
> http://www.mcs.anl.gov/~hategan/deepfieldbug.diff
> 
> (you need to apply it in swift/src/org/griphyn/vdl).
> 
> _______________________________________________
> Swift-devel mailing list
> Swift-devel at ci.uchicago.edu
> http://mail.ci.uchicago.edu/mailman/listinfo/swift-devel

-- 
Michael Wilde
Computation Institute, University of Chicago
Mathematics and Computer Science Division
Argonne National Laboratory