Hi -
I had reported bugs to do with nested records/classes that seem
similar to your problems here - those are
in test/users/ferguson/nested and were fixed last August by Tom H.
I think that you're running into a related but different problem.
It would be good to file your example (or even better several variants
of it) in the test system as a .future bug. Are you able to do so?
(see doc/developer/bestPractices/ContributorInfo.rst if you
haven't already)
I agree with Brad that you shouldn't need to add an 'outer' argument;
however the program still fails to compile if I remove that
constructor. Many simple variations of this program fail:
* removing constructor
* moving array declaration to variable in a method
* commenting out 'key' and/or 'value' from the record
-michael
On 5/5/15, 2:07 AM, "Brad Chamberlain" <br...@cray.com> wrote:
>Hi Brian --
>
>
>Just to get a response to you quickly -- I suspect that you're doing
>anything inherently wrong or off-base, but are just nudging up against
>the limit to which our current constructor and nesting features work well
>together. You may have seen elsewhere
> that our constructor design and implementation are currently in flux and
>on the way to a healthier implementation which should hopefully fix up
>things like this.
>
>
>The main part that surprised me looking at your code was the implication
>that the compiler required you to add an 'outer' argument, which is
>something I think of as being what it should be adding implicitly in the
>case of nested stuff, when needed. The
> other thing I would experiment with, if I weren't going to bed, is
>whether making the class and record non-generic simplifies things at all.
>
>
>The fact that de-nesting them makes things work is what makes me
>suspicious that it's our problem rather than yours. But nested records
>and classes do tend to throw me for a loop sometimes, especially this
>late in the evening...
>
>
>-Brad
>
>
>
>
>________________________________________
>From: Brian Guarraci [br...@ops5.com]
>Sent: Monday, May 04, 2015 9:17 PM
>To: chapel-developers@lists.sourceforge.net
>Subject: [Chapel-developers] nested classes
>
>
>Hi,
>
>
>I've run into a strange error while using I felt was a natural use of
>nested records. Below is a snippet. It tries to use the parameterized
>types to create an nested record which is then entries in a hash.
>There's a few errors when I compile and eventually
> the only way to get it to work was to pull ValueType and TableEntry up
>into the top-level of the module. Any ideas? I didn't see much on this
>in the spec.
>
>
>Thanks!
>Brian
>
>
>/*
>test/nested_class.chpl:22: error: unresolved type specifier
>'LockFreeHash(uint(32)).TableEntry'
>test/nested_class.chpl:13: note: candidates are: TableEntry(_mt: _MT,
>outer)
>*/
>module LockFreeHash {
>
> type KeyType = uint(32);
>
>
> class LockFreeHash {
> type ValueType = uint(32);
> var hashSize: uint(32) = 1024*1024;
>
>
> record TableEntry {
> var key: atomic KeyType;
> var value: atomic ValueType;
>
>
> // test/nested_class.chpl:18: error: constructor for class
>'TableEntry' requires a generic argument called 'outer'
> proc TableEntry(outer: LockFreeHash) {
> key.write(0);
> }
> }
>
>
> var array: [0..hashSize-1] TableEntry;
>
>
> // removed methods
> }
>}
>
>
>
>
>
------------------------------------------------------------------------------
One dashboard for servers and applications across Physical-Virtual-Cloud
Widest out-of-the-box monitoring support with 50+ applications
Performance metrics, stats and reports that give you Actionable Insights
Deep dive visibility with transaction tracing using APM Insight.
http://ad.doubleclick.net/ddm/clk/290420510;117567292;y
_______________________________________________
Chapel-developers mailing list
Chapel-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/chapel-developers
