View Issue Details

Related ChangesetsJump to NotesJump to History
IDProjectCategoryView StatusDate SubmittedLast Update
0024370CommunityOCCT:Shape Healingpublic2013-11-16 23:302014-05-05 13:37
ReporterRoman Lygin Assigned Tobugmaster  
PrioritynormalSeverityminor 
Status closedResolutionfixed 
PlatformAOSL 
Product Version6.7.0 
Target Version6.7.1Fixed in Version6.7.1 
Summary0024370: [Regression] 6.7.0beta ShapeFix_EdgeProjAux breaks conventions on using IsDone flag
DescriptionUp to 6.6.0 inclusive, ShapeFix_EdgeProjAux used a straightforward convention (even if not explicitly documented): if the edge's vertices cannot be projected within requested tolerance then IsDone (first and/or last) was set to false.

6.7.0 beta broke this convention and sets IsDone to true unless projection completely failed and the returned distance is infinite.
This broken logic does not allow calling algorithms to identify the case when projection failed within initially requested tolerance. Moreover returned values of parameters (FirstParam() and LastParam() now correspond to pcurve ends, not to really projected values).

Apparently regression is introduced by fixing a bug 0024206. The fix should be reworked so that the previous logic is preserved.
Steps To ReproduceSee the enclosed reproducer.
TagsNo tags attached.
Test case numberbugs heal bug24370

Attached Files

  • testcase.zip (4,450 bytes)

Relationships

parent of 0024543 closedbugmaster Open CASCADE Error in de/step_3/E6 test case for DEBUG mode 

Activities

Roman Lygin

2013-11-16 23:30

developer  

testcase.zip (4,450 bytes)

gka

2013-11-22 11:18

developer   ~0026835

I would like to note that this fix is not regression but changing behavior.
After this fix flags IsFirstDone() and IsLastDone() set to false only for cases
when projection for corresponding parameter was not made.

If projection is performed with distance more than specified precision then it is possible to get this solution and decide to use it or not.
In previous version it was necessary to repeat projection with greater value of specified precision from time to time.
Please note that class ShapeFix_EdgeProjAux is not API class.

Roman Lygin

2013-11-22 11:45

developer   ~0026839

Galina, please have a look at the reproducer.
The real projecting parameters correspond to those given in case 2: [5.2435913905776836e-005, 0.99999998584035321].
If you call an algorithm with precision of Precision::Confusion() it will return you flags as IsDone() and the range [0, 1]. In no way can you understand that actually the algorithm failed to project and you have to increase the tolerance to, say 0.2, to get a real range [5.24e-5, 0.999.584].

Previous behavior was intuitively clear and returned you IsDone false unless it could reach the requested projection. In new behavior the tolerance stops making any sense.

Note that the new behavior not just triggered regressions in CAD Exchanger. It jeopardizes own OCC STEP importer (StepToTopoDS_TranslateEdgeLoop) which relies on logic of ShapeFix_EdgeProjAux:

      myEdgePro->Init (Face, edge);
      myEdgePro->Compute(preci);
      if (myEdgePro->IsFirstDone() && myEdgePro->IsLastDone()) {
        if (Abs (myEdgePro->FirstParam() - myEdgePro->LastParam()) < Precision::PConfusion())
          continue;
        B.Range(edge, Face,myEdgePro->FirstParam(), myEdgePro->LastParam());
      }
      else {
        RemoveSinglePCurve(edge, Face);
#ifdef DEBUG
        cout <<"Removing after prj"<<endl;
#endif
      }

Where in the past, false IsDone() would cause ignoring a wrong pcurve, it will now be accepted almost certainly (given that IsDone() would return false only if projection screwed up completely). This will lead to an edge containing inconsistent 3D and 2D curves (and with incorrect ranges!) leading to more broken shape downstream.

That is why, I believe the previous logic should be preserved and it is an underlying implementation that needs to be changed.


If for whatever reason you do not want to revert the original behavior (and I don't see a strong argument why), here is an alternative:
- do not accept any user-defined tolerance
- always do projection (e.g. calling ShapeAnalysis_Curve::Project() with Precision::Infinite())
- return [square] distance from curve points to respective vertices - this will allow to take a decision of successful/failed projection by the user outside of the algorithm.

gka

2013-11-25 16:47

developer   ~0026890

Branch CR24370 is ready to be reviewed.

gka

2013-11-25 17:06

developer   ~0026893

In order to avoid removing 2D curve written in file following behavior is implemented now:
If projection of points from 3D curve corresponding range of edge was performed then obtained parameters on 2D curve are used independent from distance points from 3D curve for projected points.

gka

2013-11-26 16:24

developer   ~0026923

Branch CR24370_1 is ready to be reviewed

abv

2013-11-26 17:08

manager   ~0026926

No remarks, please test

mkv

2013-11-27 11:31

tester   ~0026946

Dear BugMaster,

Branch CR24370_1 (and products from GIT master) was compiled on Linux and Windows platforms and tested.
SHA-1: bf7b996e99ef5f40239c1152a6210fc0280efd7e

Number of compiler warnings:

occt component :
Linux: 320 (320 on master)
Windows: 0 (0 on master)

products component :
Linux: 188 (188 on master)
Windows: 286 (286 on master)

Regressions/Differences:

http://occt-tests/CR24370-1-master-occt/Debian60-64/de/iges_1/P5.html
http://occt-tests/CR24370-1-master-occt/Windows-32-VC9/de/iges_1/P5.html
de iges_1(001) P5: FAILED

http://jenkins-test-02.nnov.opencascade.com:8080/user/mnt/my-views/view/CR24370_1/job/mnt-CR24370_1-master_products_tests_linux_start/label=Debian60-64,tests_group=sat,tests_subgroup=002/2/HTML_Report/?
http://jenkins-test-02.nnov.opencascade.com:8080/user/mnt/my-views/view/CR24370_1/job/mnt-CR24370_1-master_products_tests_windows_start/label=windows_test,tests_group=sat,tests_subgroup=002/2/HTML_Report/?
sat doc_2(002) A4, A5, A9, F7, H1, I7

Testing cases:


Testing on Linux:
Total MEMORY difference: 357511528 / 358410160
Total CPU difference: 54597.54999999992 / 43957.070000000385

Testing on Windows:
Total MEMORY difference: 412851588 / 412662448
Total CPU difference: 31332.234375 / 32172.234375

There are not differences in images found by testdiff.

gka

2013-11-27 19:46

developer   ~0026968

Test cases :
de iges_1 P5,
sat doc_2 A4, A5, A9, F7, H1, I7 are not regressions.

Could you please to regenerate these tests.

mkv

2013-11-28 15:18

tester   ~0026985

Last edited: 2013-11-28 15:19

 Dear abv,
could you please review new draw-command for issue CR24370.
(See CR24370_1)

apn

2013-12-05 14:51

administrator   ~0027107

Last edited: 2013-12-05 14:52

 Dear BugMaster,

Branch CR24370_1 (and products from GIT master) was compiled on Linux and Windows platforms and tested.
SHA-1: 7ff733395e79c2b40767e6f5c38a1f2925abf7bd

Number of compiler warnings:

occt component :
Linux: 103 (103 on master)
Windows: 0 (0 on master)

products component :
Linux: 188 (188 on master)
Windows: 237 (237 on master)

Regressions/Differences:
No regressions/differences

Testing cases:
It's necessary to create test case

Testing on Linux:
Total MEMORY difference: 358781576 / 358643576
Total CPU difference: 43486.840000000055 / 45654.32000000018

Testing on Windows:
Total MEMORY difference: 413244472 / 412408396
Total CPU difference: 34745.5625 / 33977.296875

There are no differences in images found by testdiff.

apn

2013-12-05 14:53

administrator   ~0027108

PRODUCTS branch CR24370_1 (contains changes in reference data in test cases sat/doc_2/A4 A5 A9 F7 H1 was created.

mkv

2014-01-15 11:43

tester   ~0027490

Dear BugMaster,

Test case (bugs/heal/bug24370) is bad (input shape are bad)

pload QAcommands
OCC24370 bug24370_edge.brep bug24370_pcurve_draw bug24370_surface_draw 0.001

OCC24370 Eror: Null input edge

mkv

2014-01-15 15:40

tester   ~0027498

Dear BugMaster,

Test case bugs/heal/bug24370 was corrected.
It is OK now.
Branch CR24370_1 is TESTED.

Related Changesets

occt: master e3096dec

2014-01-16 09:11:45

gka


Committer: bugmaster Details Diff		 0024370: [Regression] 6.7.0beta ShapeFix_EdgeProjAux breaks conventions on using IsDone flag.

Projection of 3d points corresponding to range of edge on 2d curve considered as successful for all cases to except for cases when projection is not made.

Initialization initial values of flags
Modification initial values of flags setting status Done for first and last parameters

Adding new draw-command for issue CR24370

Test case for issue CR24370		 Affected Issues
0024370
mod - src/QABugs/QABugs_19.cxx Diff File
mod - src/ShapeFix/ShapeFix_EdgeProjAux.cxx Diff File
add - tests/bugs/heal/bug24370 Diff File
mod - tests/de/step_3/E6 Diff File
mod - tests/de/step_4/C9 Diff File
mod - tests/de/step_4/D1 Diff File

Issue History

Date Modified Username Field Change
2013-11-16 23:30 Roman Lygin New Issue
2013-11-16 23:30 Roman Lygin Assigned To => gka
2013-11-16 23:30 Roman Lygin File Added: testcase.zip
2013-11-21 18:29 gka Status new => assigned
2013-11-22 11:18 gka Note Added: 0026835
2013-11-22 11:18 gka Assigned To gka => Roman Lygin
2013-11-22 11:18 gka Status assigned => feedback
2013-11-22 11:45 Roman Lygin Assigned To Roman Lygin => gka
2013-11-22 11:45 Roman Lygin Note Added: 0026839
2013-11-22 11:45 Roman Lygin Status feedback => assigned
2013-11-25 16:47 gka Note Added: 0026890
2013-11-25 16:47 gka Assigned To gka => abv
2013-11-25 16:47 gka Status assigned => resolved
2013-11-25 17:06 gka Note Added: 0026893
2013-11-26 16:24 gka Note Added: 0026923
2013-11-26 17:08 abv Note Added: 0026926
2013-11-26 17:08 abv Assigned To abv => bugmaster
2013-11-26 17:08 abv Status resolved => reviewed
2013-11-27 11:31 mkv Note Added: 0026946
2013-11-27 11:32 mkv Assigned To bugmaster => gka
2013-11-27 11:32 mkv Status reviewed => assigned
2013-11-27 19:46 gka Note Added: 0026968
2013-11-27 19:46 gka Status assigned => resolved
2013-11-27 19:47 gka Status resolved => reviewed
2013-11-28 15:18 mkv Note Added: 0026985
2013-11-28 15:19 mkv Assigned To gka => abv
2013-11-28 15:19 mkv Status reviewed => feedback
2013-11-28 15:19 mkv Note Edited: 0026985
2013-11-28 17:16 abv Assigned To abv => gka
2013-12-04 10:13 gka Status feedback => reviewed
2013-12-04 11:41 mkv Assigned To gka => apn
2013-12-05 13:02 apn Assigned To apn => mkv
2013-12-05 14:51 apn Note Added: 0027107
2013-12-05 14:52 apn Note Edited: 0027107
2013-12-05 14:53 apn Note Added: 0027108
2013-12-05 14:53 apn Status reviewed => feedback
2013-12-21 10:13 abv Target Version 6.7.0 => 6.7.1
2014-01-15 11:43 mkv Note Added: 0027490
2014-01-15 11:43 mkv Assigned To mkv => abv
2014-01-15 15:40 mkv Note Added: 0027498
2014-01-15 15:40 mkv Test case number => bugs heal bug24370
2014-01-15 15:41 mkv Assigned To abv => bugmaster
2014-01-15 15:41 mkv Status feedback => tested
2014-01-21 10:41 mkv Relationship added parent of 0024543
2014-01-21 11:22 bugmaster Changeset attached => occt master e3096dec
2014-01-21 11:22 bugmaster Status tested => verified
2014-01-21 11:22 bugmaster Resolution open => fixed
2014-05-05 13:36 aiv Status verified => closed
2014-05-05 13:37 aiv Fixed in Version => 6.7.1