[6.8.0beta regression] gp_Trsf2d::SetMirror() looses transformation form

The regression is due to some recent commits in master. Most likely - #0025194, where SetValues() is called at the end of SetMirror() and sets the form to gp_CompoundTrsf.

Please double-check all the methods where SetValue() is now called, both in gp_Trsf2d and gp_Trsf. Adding unit tests for all those methods (e.g. as simple as the reproducer above) would be useful.

A quick fix to set shape form after calling SetValues() did the trick and restored previous behavior:

void gp_Trsf2d::SetMirror (const gp_Ax2d& A)
{
// shape = gp_Ax1Mirror; //RLN
  const gp_Dir2d& V = A.Direction ();
  const gp_Pnt2d& P = A.Location ();
  Standard_Real VX = V.X();
  Standard_Real VY = V.Y();
  Standard_Real X0 = P.X();
  Standard_Real Y0 = P.Y();

  SetValues(1.0-2.0*VX*VX, -2.0*VX*VY, -2.0*((VX * VX - 1.0)*X0 + (VX*VY*Y0)),
            -2.0*VX*VY, 1.0-2.0*VY*VY, -2.0*((VX*VY*X0)+(VY*VY-1.0)*Y0));

  shape = gp_Ax1Mirror; //RLN
  scale = - 1.0;
}

To make sure the fix is complete in other places I will refrain from pushing the fix. Instead I'd rather suggest the author of #0025194 to make another review and another commit to address this regression.

Branch [archived branch] has been created by Participant.

[revision removed]


Detailed log of new commits:

Author: nbv
Date: Thu Oct 16 10:27:53 2014 +0400

    0025374: [6.8.0beta regression] gp_Trsf2d::SetMirror() looses transformation form
    
    1. Method Orthogonalize is called in case of transform matrix change.
    2. gp_TrsfForms are set after SetValues(...) method calling from methods of gp_Trsf2d class.

Dear Andrey!

Please review branch CR25374.

Branch [archived branch] has been updated forcibly by Participant.

[revision removed]

Dear Andrey!

Please review CR25374 branch again.

No remarks, please test

Branch [archived branch] has been updated forcibly by Participant.

[revision removed]

Dear Commenter 1,

Branch CR25374 (and products from GIT master) was compiled on Linux, MacOS and Windows platforms and tested.
[revision removed]

Number of compiler warnings:
occt component:
   Linux: 15 (15 on master)
   Windows: 0 (0 on master)
products component:
   Linux: 11 (11 on master)
   Windows: 1 (1 on master)

Regressions/Differences:
http://occt-tests/CR25374-master-occt/Debian60-64/summary.html
http://occt-tests/CR25374-master-occt/Windows-32-VC10/summary.html
de iges_1(001) J9,K3
de iges_2(002) B8, E6, G1

Testing cases:
Not needed
However, necessity to create test-cases will be defined in frame of #0025389

Testing on Linux:
Total MEMORY difference: 367272704 / 367465072
Total CPU difference: 47379.94999999994 / 44339.100000000006

Testing on Windows:
Total MEMORY difference: 255002692 / 252142604
Total CPU difference: 32317.875 / 34651.0625

There are differences in images found by testdiff:
http://occt-tests/CR25374-master-occt/Debian60-64/diff-Debian60-64.html
http://occt-tests/CR25374-master-occt/Windows-32-VC10/diff-Windows-32-VC10.html
Pay attention to bugs iges bug13627 (on Windows only)

Dear Alexei!

Please update regression tests according to their new behavior.

The bug should be TESTED.

P.S.
The solution about test case will be applied after the fix of bug #0025389 related.

Branch [archived branch] has been updated by Participant.

[revision removed]


Detailed log of new commits:

Author: apv
Date: Thu Oct 23 17:19:21 2014 +0400

    Update of test-cases

Test-cases are updated.

Note that update of DE test cases for this issue mostly reverts changes made in #0025194

The reason of these changes was in forceful orthogonalize of transformation matrix for corresponding transformations (mirror, translation etc.).

Now these matrix are not orthogonalized because they are considered to be orthogonal. However, this condition is satisfied with some error, which are decreased after forceful orthogonalize.

P.S.

The error even in ~ 1.0e-16 results in changing in shape (it will get another structure).

Branch [archived branch] has been deleted by Participant.

[revision removed]