PR 255185 Feature Set Filter Crashes PCD, application error

huangyhg

Feature Set Filter Crashes PCD, application error
<< Bob Hospadaruk  --  05/27/08  16:20:20>>
Use the attache program.  Insert a flister set right after SCN36.
Create gauss 50 upr and outlier on.  Select SCN16,20,23,26,29,32,35,36.  Create
Application error is your reward...
Critical for PT demo in progress.
<<END>>


<< Changes made by Neil Kay -- 02/24/09  13:31:37>>
Status:  RESOLVED to CLOSED
<<END>>
<< Dan Zwick  --  06/07/08  12:42:04>>
I uploaded to 4.31, 4.3B, and 4.4B as well.
<<END>>
<< Dan Zwick  --  06/07/08  08:34:04>>
The problem Andy mentioned is a new problem I found when testing the change I made for this PR. I increased the maximal number of inputs to the filter from 1 to 100 since the output for each input is a new filter set. If they really want more than 100 filter sets as output they can do it again. In testing Paola's changes I discovered the new bug, but it is unrelated to Paola's changes, which seem to work fine. The bug we found is probably related to another PR I'm working on, but I can't check the PR number while this dialog is open. Anyway, I uploaded
V43R\MENU\FILTERDL.CPPP
and will upload to the other versions when I have checked that they build.
<<END>>
<< Changes made by Dan Zwick -- 06/07/08  08:35:14>>
Action:  Dan Zwick to Bob Hospadaruk, Status:  OPEN to RESOLVED
<<END>>
<< Andy Roberts  --  06/06/08  17:52:04>>
Dan and I see that the problem is that we are sending an initial vector of (0,0,0) for the cylinder solution.  We are working on a fix.
<<END>>
<< Paola Pallo  --  06/06/08  16:41:15>>
Dan, from your comment in 255186, these changes should be backed out. To avoid futher confusion,  I think it would be better you do it personally.
Backing them out ,PC-DMIS immediately answers to Bob's enhancement request ( multiple separate filter set commands from the same dialog) . Thanks
<<END>>
<< Changes made by Paola Pallo -- 06/06/08  16:41:40>>
Action:  Bob Hospadaruk to Dan Zwick, Status:  RESOLVED to OPEN
<<END>>
<< Dan Zwick  --  05/28/08  10:07:54>>
With Bill's permission I have uploaded:
V43R\INCLUDE\FILTERDL.H
V43R\MENU\FILTERDL.CPP
V431\INCLUDE\FILTERDL.H
V431\MENU\FILTERDL.CPP
V44B\INCLUDE\FILTERDL.H
V44B\MENU\FILTERDL.CPP
<<END>>
<< Bill Wilcox  --  05/28/08  16:29:33>>
Dan please add your fix to V43R and V431 as well as V44.
<<END>>
<< Changes made by Bill Wilcox -- 05/28/08  16:29:35>>
Priority:  Critical to Stop Rel.
<<END>>
<< Bill Wilcox  --  05/28/08  16:26:53>>
Hi Bob, please log enhancment request in as enhancments. Your comment will not get noticed or acted upon if you add it to a problem report.
<<END>>
<< Dan Zwick  --  05/28/08  08:13:36>>
I've fixed the crash by only allowing one input from the filter dialog box. It's already not possible to enter more than one input from the edit window. A filter set is a constructed feature and the current design only envisions a single output. If you want to treat all of your scans as a single set, that is your prerogative. You can already do that by first constructing a set containing the scans and then using that set as input to the filter set. Just don't expect it to work particularly well in your case, for the reasons I already mentioned.
Files uploaded:
V43B\INCLUDE\FILTERDL.H
V43B\MENU\FILTERDL.CPP
<<END>>
<< Changes made by Dan Zwick -- 05/28/08  08:14:59>>
Action:  Dan Zwick to Bob Hospadaruk, Status:  OPEN to RESOLVED
<<END>>
<< Bob Hospadaruk  --  05/28/08  10:42:13>>
Please do fix the crash although I don’t understand the design. A filter set is a kind of feature set, that is, a set of points.  And you can have multiple inputs into a feature set, and you allow such a feature set as input to a filter set, so why not allow the selection of multiple inputs to create one set that is filtered.
If we won’t allow multiple inputs, then I hope the dialog will still allow multiple selections, but then create multiple separate filter set commands
<<END>>
<< Changes made by Tim Wernicke -- 05/27/08  18:21:11>>
Action:  Tim Wernicke to Dan Zwick, Assigned:   to Dan Zwick, Priority:  to Critical
<<END>>
<< Dan Zwick  --  05/27/08  15:54:37>>
Bob, the input to a filter set is a single feature, not multiple features. It can be a scan comprising several basic scans. In that case, each of the basic scans is filtered individually. It can also be a filter set (in case you want to filter the data again, possibly with a different filter) or a constructed set. If the input is a constructed set then it is treated as one big scan. For example, you could create a set from scans 16, 20, 23, 26, etc., and then filter it. However, the Gauss filter expects the data to be one complete circle without gaps, so this might not work very well. A cylindrical filter might work better, but there would still be gaps where one scan ends and the next starts. Of course, you could just filter each of the scans separately. A cylindrical filter would be better for those scans that only span a partial circle.
This doesn't mean that it's ok for PC-DMIS to crash if you select multiple inputs. I'll leave this PR open until we've fixed that.
<<END>>
<< Dan Zwick  --  05/27/08  14:06:14>>
I'll take a look.
<<END>>