Thanks for coming back to me about this.
I have spent the day tinkering with this. There does seem to be some
problem when destructing a view which contains sub-views. It certainly
seems like memory is leaked - have you seen that at all ? I think
constructing / destructing a view (containing sub-views) would be a good
start.
Any pointers as to what might be wrong would be much appreciated.
Best regards,
Ian Fairclough.
Ian,
> The view does contain a number of subviews. The data file is 72M and
> the
> initial view contains 28104 rows.
The presence of subviews may be important.
> #define d4_MetaKitLibraryVersion 249 // 2.4.9.2 release, Mar 17,
> 2003
Good.
> I have built the "thread safe" version.
Can you test with a non-threaded build? I doubt that it matters, but
if all else fails it will be important to verify this.
> Any suggestions for things I might try would be much appreciated.
Best thing would be a debug build and interrupting the hanging process
to see where it is hanging exactly... i.e. a stack trace.
A reproducible test case would be good as well, since then I can do
such checks myself. I doubt that the problem is data related, so
actual contents in such a test can probably easily be generated.
-jcw
----- Original Message -----
From: "Ian Fairclough" <[EMAIL PROTECTED]>
To: <[EMAIL PROTECTED]>
Cc: <[EMAIL PROTECTED]>
Sent: Friday, October 31, 2003 7:22 PM
Subject: Lock-up in view constructor
> To answer the questions :
>
> The view does contain a number of subviews. The data file is 72M and the
> initial view contains 28104 rows.
>
> Metakit version is :
>
> #define d4_MetaKitLibraryVersion 249 // 2.4.9.2 release, Mar 17, 2003
>
> I have built the "thread safe" version.
>
> Nothing jumps out at me from the release notes. I see a comment about a
> race condition in the C++ build (2003-03-05 Fixed two potential races
in
> C++ threaded build) but it looks like that was included in 2.4.9.2.
>
> Any suggestions for things I might try would be much appreciated.
>
> Best regards,
> Ian Fairclough.
>
> > in someClass constructor :
> > view = storage.GetAs(structure);
> >
> > void someClass::dummyFunction
> > {
> > c4_View selectedDataView = this->view.duplicate();
> > }
> >
> > My problem occurs when the "dummyFunction" is called and returns i.e.
> > when
> > the destructor for "selectedDataView" is called.
>
> Weird. Does "view" contain subviews?
>
> > Is auto commit turned on by default ?
>
> Not in C++ and Python.
>
> >>> I am running under Red Hat Linux 7.3. Metakit was built with gcc
> >>> version
> >>> 2.96 20000731.
>
> What is the version of MK - see #define d4_MetaKitLibraryVersion in
> "mk4.h".
>
> You can then compare this with the current change log:
> http://www.equi4.com/metakit/viewcvs.cgi/metakit/CHANGES?rev=1.172
>
_______________________________________________
metakit mailing list - [EMAIL PROTECTED]
http://www.equi4.com/mailman/listinfo/metakit
