Well, I was almost right the first time. When you get a new invalid
positions component, you may invalidate a different set of connections, and
therefore invalidate still more positions because there are no connections
elements incident on them. (It may seem odd that unreferenced positions
are considered invalid, but that was an architectural decision made long
ago). However, if you have no connections component, then you have
isolated points and the lack of incident positions does not invaliate them.
We should have required a connections component of element type "points" to
be more consistent, but we didn't. In this case, I needed to add code to
keep DXInvalidateConnections and DXInvalidateUnreferencedPositions if there
were no connections. Sorry 'bout that.
Greg
Gregory D
Abram/Watson/[EMAIL PROTECTED] To:
[email protected]
Sent by: cc:
[EMAIL PROTECTED] Subject: Re: [opendx-users] Re:
Invalid positions- looks like a bug
son.ibm.com
08/18/2003 03:06 PM
Please respond to
opendx-users
Looking into it right now...
Greg
Chris Pelkie
<[EMAIL PROTECTED]> To:
[email protected]
Sent by: cc:
[EMAIL PROTECTED] Subject: Re:
[opendx-users] Re: Invalid positions- looks like a bug
son.ibm.com
08/18/2003 07:41 AM
Please respond to
opendx-users
Yeah, I remember that, came across it before posting, but didn't
realize the significance til now.
Let me see if I can rework the way I was doing things to see if Greg's
fix solved the underlying problem that led me to do it that way.
The original problem being solved with my method was to enable proper
OR'ing of two 'invalids' components to make a joint invalids. I seem to
remember I had to play around with Rename and Remove and Replace to
overcome problems with fields that had prior invalids not behaving as
expected during downstream ops that then ignored or bellyached about
those invalids. But the details are sketchy so I have to review the net
(which has some ungodly number of pages, like 35-40). I'll followup
later.
On Sunday, Aug 17, 2003, at 12:15 America/New_York, David Thompson
wrote:
> Chris,
>
> The problem is due to a fix that Greg had added in helper.c (see
> http://opendx.npaci.edu/mail/opendx-dev/2002.09/msg00009.html). I
> backed this fix out, and it goes back to working as expected.
>
> Greg, will you have time to look at this, or should I see what I can
> do?
>
________________________________
Chris Pelkie
Vice President
Conceptual Reality Presentations, Inc.
30 West Meadow Drive
Ithaca, NY 14850
