View Issue Details [ Jump to Notes ] [ Print ]
IDProjectCategoryView StatusDate SubmittedLast Update
0006970ParaView(No Category)public2008-05-06 22:572009-05-13 13:59
ReporterAlan Scott 
Assigned ToBurlen 
PriorityimmediateSeveritycrashReproducibilityalways
StatusclosedResolutionfixed 
PlatformOSOS Version
Product Version 
Target Version3.4Fixed in Version3.4 
Summary0006970: (MARS - immediate) Extract CTH Boundaries and Fragments crash - multiple servers
DescriptionWe have a bug in Extract CTH Bounaries and Fragments, when using client/server and using 4 pvservers. This bug does not appear with 2 pvservers. I wonder if it is related to bug number 6969.

* Start paraview, client/server, linux/linux. Use 4 pvservers.
* Open dave's CTHExamples/cth-med/spcta.[0-31]. Turn on all cell arrays. Apply. This will load properly.
* Extract CTH Boundaries and Fragments.
  - Select Material volume fraction - 1. Apply. It works.
>>> Select Material Volume fraction - 2. The server never returns with 4 servers. It returns very quickly if we only have 2 servers. This is extremely repeatable.
 
TagsNo tags attached.
Project
Topic Name
Type
Attached Files

 Relationships

  Notes
(0011728)
Alan Scott (manager)
2008-05-06 22:59

 Woops - I forgot. This bug was probably introduced sometime last week, maybe late April. It was in the code (and mistakenly thought to be a compiler bug) before the checkin on May 6.
(0011729)
Alan Scott (manager)
2008-05-06 23:06
edited on: 2008-05-06 23:06

 One more note, a stack trace, that may give a hint...

On node 0, also another node:

in btl_openib_component_progress()
in mca_bml_r2_progress()
in opal_progress()
in mca_pml_ob1_recv()
in PMPI_Recv
in vtkMPICommunicatorReceiveData<char>

Appears to be stuck there...
(0012353)
Alan Scott (manager)
2008-06-16 19:54

 Sorry, it still is bad. Now we have a crash.
* Start Paraview, client/server, XP/linux, 4 pvservers.
* Open dave's CTHExamples/cth-med/spcta.[0-31]. Turn on all cell arrays. Apply.
* Extract CTH Boundaries and Fragments.
  - Select Material volume fraction - 2, on Select Mass Arrays only use Mass -1, Compute volumen weighted average over: all except the two masses. Apply. It works.
>>>> Rotate the object. Server dies.

Utkarsh, I am setting this to you since you marked it as resolved. Is that correct?
(0012356)
Alan Scott (manager)
2008-06-16 20:20

 I have a bad build. Need to retest.
(0012364)
Alan Scott (manager)
2008-06-17 12:16

 This is a crash. See my notes from 6-16.
(0012375)
Burlen (developer)
2008-06-17 14:36

 Process id: 2 >> ERROR: In /ext/kitware_cvs/ParaView3/Servers/Filters/vtkCTHFragmentConnect.cxx, line 4627
vtkCTHFragmentConnect (0x8d14710): Unexpected edge case:14

Process id: 2 >> ERROR: In /ext/kitware_cvs/ParaView3/Servers/Filters/vtkCTHFragmentConnect.cxx, line 4627
vtkCTHFragmentConnect (0x8d14710): Unexpected edge case:14

By the way, a usage note(Does not affect the bug). The material fraction and mass arrays are paired. What this means is that you will get a garbage result if you use the wrong ones together. For example if you want to generate fragments for Material fraction 1 you need to use Mass-1, if you want fragments for material 2 you need to use Mass-2. In either case I see the edge case.
(0012574)
Burlen (developer)
2008-06-30 10:32

 Charles looked at it, and noted that the neighbors were being incorrectly generated...
(0013056)
Alan Scott (manager)
2008-08-18 22:00

 Tested client/server. Although it seems to always work, I had one crash (after running other bad stuff earlier) that had to do with a material fraction of 1 and a Mass arrays of 2. This is a no-no, and I could not repeat the crash. Until we can replicate it, lets mark this as closed.

 Issue History
Date Modified Username Field Change
2008-05-06 22:57 Alan Scott New Issue
2008-05-06 22:59 Alan Scott Note Added: 0011728
2008-05-06 23:06 Alan Scott Note Added: 0011729
2008-05-06 23:06 Alan Scott Note Edited: 0011729
2008-05-13 12:43 Ken Moreland Status backlog => tabled
2008-05-13 12:43 Ken Moreland Assigned To => Berk Geveci
2008-06-03 11:18 Utkarsh Ayachit Status tabled => @80@
2008-06-03 11:18 Utkarsh Ayachit Resolution open => fixed
2008-06-16 19:54 Alan Scott Assigned To Berk Geveci => Utkarsh Ayachit
2008-06-16 19:54 Alan Scott Status @80@ => @20@
2008-06-16 19:54 Alan Scott Resolution fixed => reopened
2008-06-16 19:54 Alan Scott Note Added: 0012353
2008-06-16 20:19 Alan Scott Status @20@ => tabled
2008-06-16 20:19 Alan Scott Assigned To Utkarsh Ayachit => Alan Scott
2008-06-16 20:20 Alan Scott Status tabled => @80@
2008-06-16 20:20 Alan Scott Resolution reopened => fixed
2008-06-16 20:20 Alan Scott Note Added: 0012356
2008-06-17 12:16 Alan Scott Assigned To Alan Scott => Utkarsh Ayachit
2008-06-17 12:16 Alan Scott Status @80@ => @20@
2008-06-17 12:16 Alan Scott Resolution fixed => reopened
2008-06-17 12:16 Alan Scott Note Added: 0012364
2008-06-17 12:24 Utkarsh Ayachit Status @20@ => tabled
2008-06-17 12:24 Utkarsh Ayachit Assigned To Utkarsh Ayachit => Burlen
2008-06-17 14:36 Burlen Note Added: 0012375
2008-06-30 10:32 Burlen Note Added: 0012574
2008-08-14 13:39 Ken Moreland Status tabled => @80@
2008-08-14 13:39 Ken Moreland Resolution reopened => fixed
2008-08-18 22:00 Alan Scott Status @80@ => closed
2008-08-18 22:00 Alan Scott Note Added: 0013056
2009-05-13 13:58 Utkarsh Ayachit Target Version => 3.4
2009-05-13 13:59 Utkarsh Ayachit Fixed in Version => 3.4
2011-06-16 13:10 Zack Galbreath Category => (No Category)

Copyright © 2000 - 2018 MantisBT Team