LArSoft

About the face center, note also the documentation at http://nusoft.fnal.gov/larsoft/doxsvn/html/classgeo_1_1GeometryCore.html#a424a92278ca48cc9457e7f7ac1ab9a97 .
Also note the comment that informally deprecates DetHalfWidth(), DetHalfHeight() and DetLength(), which simply do not do what they promise.

The dump of LArIAT geometry reports:

Detector lariat has 1 cryostats:
  cryostat #0 at { 23.75; 0.2; 45 } cm has 1 TPC(s):
    TPC C:0 T:0 (49.97 x 50 x 100) cm^3 at { 24.185; -7.21645e-16; 45 }
    drift direction { -1; 0; 0 } from cathode around { 47.9; -7.21645e-16; 45 } through 48.3 cm toward 2 wire planes
    maximum wires on any plane: 240
    active volume (47.5 x 40 x 90) cm^3, main directions: width { 1; 0; 0 } height { 0; 1; 0 } length { 0; 0; 1 }
    bounding box: ( -0.8, -25, -5 ) -- ( 49.17, 25, 95 )
    active volume box: ( 0.4, -20, 0 ) -- ( 47.9, 20, 90 )
      plane C:0 T:0 P:0 at (-5.52336e-15, -7.21645e-16, 45) cm, theta: 1.0472 rad
          normal to wire: -0.523599 rad, with orientation vertical, has 240 wires measuring U with a wire pitch of 0.4 cm
          normal to plane: (1, 0, 0), direction of increasing wire number: (0, -0.5, 0.866025) [wire frame normal: (1, 0, 0)] (increases with z)
          wire direction: (0, 0.866025, 0.5); width 100 cm in direction: (0, 0, 1), depth 50 cm in direction: (0, -1, 0) [normal: (1, 0, 0)]
          wires cover width -43.8044 to 44.5356, depth -19.3097 to 19.7318 cm
          bounding box: ( -0.4, -25, -5 ) -- ( 2.22045e-16, 25, 95 )
      plane C:0 T:0 P:1 at (-0.4, -7.21645e-16, 45) cm, theta: 2.0944 rad
          normal to wire: 0.523599 rad, with orientation vertical, has 240 wires measuring V with a wire pitch of 0.4 cm
          normal to plane: (1, 0, 0), direction of increasing wire number: (0, 0.5, 0.866025) [wire frame normal: (1, 0, 0)] (increases with z)
          wire direction: (0, 0.866025, -0.5); width 100 cm in direction: (0, 0, -1), depth 50 cm in direction: (0, 1, 0) [normal: (1, 0, 0)]
          wires cover width -43.8044 to 44.5356, depth -19.7318 to 19.3097 cm
          bounding box: ( -0.8, -25, -5 ) -- ( -0.4, 25, 95 )
There are 0 auxiliary detectors:
End of detector lariat geometry.

The coordinates of the TPC box are: ( -0.8, -25, -5 ) - ( 49.17, 25, 95 ), size is 49.97 x 50 x 100, and the center is ( 24.185; ~0; 45 ) (all in centimetres).
The face of the TPC should be, according to these figures, 49.97 x 50 large and centered at ( 24.185, ~0, -5 ).

Considering the active volume, which is probably more interesting for physics, its coordinates include from ( 0.4, -20, 0 ) to 47.9, 20, 90 , size is 47.5 x 40 x 90, and the center is ( 23.75, 0, 45 ).
The face of the active TPC volume should be, according to these figures, 47.5 x 40 large and centered at ( 23.75, 0, 0 ).

William Foreman wrote:

In lariatsoft v06_54_00, I am noticing two oddities related to the geometry.

1) The front face centerpoint is reported incorrectly as dX/4 instead of dX/2. For example:
[...]

Result:[...]

We expect the centerpoint to be at X = 47.5/2 = 23.75 cm, but it is 11.875 cm.

Yes. That's because the implementation of geo::GeometryCore::GetTPCFrontFaceCenter() is a distillate of crap.
For the resolution I am reimplementing it.

2) The min and max X returned by the geo::TPCGeo object does not match expectations from DetWidth:
[...]

Result:
-0.8 49.17

This should be 0 and 47.5, right?

The coordinates you report are from the TPC volume, while the width you report is from the active part of it.
For the resolution, I am updating the documentation so that it explicitly mentions that DetWidth() refers to the active volume of the requested TPC.

The changes are now in LArSoft v06_59_00 .