GenFit tracking evidently needed to compare with standard Sti and Stv results.
Our standard QA application used fitted errors in DCA representation.
New conversion GenFitErrs ==> Dca errors was developed for it.

Class GFull is filled by GenFit parameters and error matrix.

Parameters:
========
Pos[3] - X,Y,Z of start of track
UVN[3][3] - three vectors (directions) defining the fitting plane

qPinv - charge/P
Uc  - cosine (track direction,UVN[0])  
Vc  - cosine (track direction,UVN[1])
U   -  projection (track space shift fitted ,UVN[0])
V   -  projection (track space shift fitted ,UVN[1])

Nc  - cosine (track direction,UVN[2]) is calculated sqrt(1-Uc*Uc-Vc*Vc)
N   -  projection (track space shift fitted ,UVN[2]) == 0 (fit is in UV plane )

Errors:
====
Triangle error matrix
<qPinv*qPinv>
<qPinv*Uc>     <Uc*Uc>
<qPinv*Vc>     <Uc*Vc> <Vc*Vc>
<qPinv*U >      <Uc*U>  <Vc*U > < U*U >
<qPinv*U >      <Uc*V>  <Vc*V > < U*V > < V*V >

When GFull is created class StDcaGenFit is filled by it
The definition of it is in $STAR/StRoot/StarRoot/GFull.h and GFull.cxx
How ro run these classes described in GFull.h

There is a test routine  GFull::TestConvertionErrs()
In this test randomly created GFull and StDcaGenFit classes
Then set of random GFull parameters is created according to given errors.
These parameters converted into Dca representation and evaluated Dca error matrix.
Then we can compare evaluated and analiticaly calculated error matrices.

This is the input distribution for GenFit. 
If name is like _A_ it is distribution of (A/sqrt(<A*A>) Mean = 0 and Rms = 1
Whe name is AB  it is distribution of (A*B - <A*B>)/sqrt(<A*A>*<B*B>) Mean = 0 and Rms <=1

Next distribution for converted to DcaErrs: If it is correct, Mean MUST be 0 and Rms for _A_ == 1

You see result is positive.

Is is following all other histograms to see that all of them are correct as well;

Input:
====

And the Output:
==========