Closed phma closed 8 years ago
The bug may not have anything to do with trianglecontours. The test fails if run by itself, but succeeds if run in gdb. Running bezitest with no arguments fails, but running all the tests individually (make test) succeeds.
Running "bezitest trianglecontours" succeeds. Running "bezitest trianglecontours closest" fails in trianglecontours.
Changing findnocubedir to use Brent's method makes this work right. There may still be an uninitialized variable lurking in the code.
When doing the trianglecontours test on Raspberry Pi, bezitest fails after saying "slope", saying that a critical point is of type 6 (PT_GRASS). The next test after slope is table, for which PT_GRASS makes sense, but it should not be finding critical points in the table.