UMLClass memory management mess (was: Dia ChangeLog report for 2006-01-13 04:00:00 UTC (Fri 13 Jan))

[Hans Breuer <hans breuer org>]

On 13.01.2006 05:20, Dia ChangeLog Daemon wrote:
> --- ChangeLog.previous  2006-01-09 05:00:46.000000000 +0100
> +++ dia-cvs-snapshot/ChangeLog  2006-01-13 05:00:05.813385280 +0100
> @@ -1,3 +1,20 @@
> +2005-01-12  Lars Clausen  <lars raeder dk>
> +
> +       * objects/UML/umlattribute.c:
> +       * objects/UML/class_dialog.c: Better way of handling
> +       connectionpoints: Don't copy them, just reference them.  Finally
> +       fixes bug #314153 (not quite the pi bug:)
> +

Hi Lars et al.,
your recent change in object/UML/class_dialog.c and the change
to umlattr1bute.c (missing a ChangeLog) entry introduces an
inconsitency again with regard to the memory management of
connections.
While before your change both uml_attribute_new() and
uml_operation_new() did create their connection points themself
now there was uml_attribute_new()->left_connection == NULL but
uml_operation_new()left_connection != NULL.
Also the connections were only sometimes initialized after 
uml_attribute_new(), which is obviously wrong.

Although I was the one moving the connection memory management
down to UMLAttribute and UMLOperation I've now reverted this back
to how it was up to until 0.94.
Although the scheme is rather complicate it appears to work again.
It seems as if this is really fixing bug #314153 and #326453. At
least as far as I've tested (connecting, deleting,
attributes/operations, undo, redo, ...)

I'd like if you could take a look and/or give it a try. It is
just commited to cvs.

Thanks,
        Hans

-------- Hans "at" Breuer "dot" Org -----------
Tell me what you need, and I'll tell you how to
get along without it.                -- Dilbert