|
| 1 | +##### |
| 2 | +Debug |
| 3 | +##### |
| 4 | + |
| 5 | +There are several situations in which you may need to enable or |
| 6 | +configure debug with PyVSC. The most common is when a set of |
| 7 | +constraints fails to solve, and diagnostics must be enabled |
| 8 | +to help understand the reason for the failure. PyVSC targets |
| 9 | +execution speed over verbosity, so default behavior is to |
| 10 | +create no diagnostics when a solve failure occurs. |
| 11 | + |
| 12 | +Enabling Solve-Fail Debug |
| 13 | +========================= |
| 14 | + |
| 15 | +PyVSC provides an optional argument to the `randomize` and |
| 16 | +`randomize_with` method to enable solve-fail debug on a |
| 17 | +per-call basis. |
| 18 | + |
| 19 | +.. code-block:: python3 |
| 20 | +
|
| 21 | + class my_e(IntEnum): |
| 22 | + A = auto() |
| 23 | + B = auto() |
| 24 | + C = auto() |
| 25 | + |
| 26 | + @vsc.randobj |
| 27 | + class my_c(object): |
| 28 | + |
| 29 | + def __init__(self): |
| 30 | + self.e = vsc.rand_enum_t(my_e) |
| 31 | + self.a = vsc.rand_uint8_t() |
| 32 | + |
| 33 | + @vsc.constraint |
| 34 | + def a_c(self): |
| 35 | + self.a == 1 |
| 36 | + with vsc.if_then(self.a == 2): |
| 37 | + self.e == my_e.A |
| 38 | + |
| 39 | + it = my_c() |
| 40 | +
|
| 41 | + with it.randomize_with(solve_fail_debug=1): |
| 42 | + it.a == 2 |
| 43 | +
|
| 44 | +In the example above, the class-level constraint set |
| 45 | +forces a==1, while the user's inline constraints forces |
| 46 | +a==2. The `randomize_with` call sets solve_fail_debug=1, |
| 47 | +which triggers creation of diagnostic information when |
| 48 | +a solve failure occurs. |
| 49 | + |
| 50 | +In this case, the output is of the following form: |
| 51 | + |
| 52 | +.. code-block:: |
| 53 | +
|
| 54 | + Problem Set: 2 constraints |
| 55 | + <unknown>: |
| 56 | + (a == 1); |
| 57 | + <unknown>: |
| 58 | + (a == 2); |
| 59 | + |
| 60 | +Enabling solve-fail debug can also be enabled globally |
| 61 | +by calling `vsc.vsc_solvefail_debug(1)` from Python code. |
| 62 | +The environment variable VSC_SOLVEFAIL_DEBUG can also |
| 63 | +be set to 1. |
| 64 | + |
| 65 | +Capturing Source Information |
| 66 | +============================ |
| 67 | +Note that no source information is available for the constraints. |
| 68 | +This is because querying source information in Python is quite |
| 69 | +time-consuming. |
| 70 | + |
| 71 | +Enabling the capture of source information for constraints can be |
| 72 | +done in two ways: via the `randobj` decorator, and via an |
| 73 | +environment variable. |
| 74 | + |
| 75 | +.. code-block:: python3 |
| 76 | +
|
| 77 | + class my_e(IntEnum): |
| 78 | + A = auto() |
| 79 | + B = auto() |
| 80 | + C = auto() |
| 81 | + |
| 82 | + @vsc.randobj(srcinfo=True) |
| 83 | + class my_c(object): |
| 84 | + |
| 85 | + def __init__(self): |
| 86 | + self.e = vsc.rand_enum_t(my_e) |
| 87 | + self.a = vsc.rand_uint8_t() |
| 88 | + |
| 89 | + @vsc.constraint |
| 90 | + def a_c(self): |
| 91 | + self.a == 1 |
| 92 | + with vsc.if_then(self.a == 2): |
| 93 | + self.e == my_e.A |
| 94 | + |
| 95 | + it = my_c() |
| 96 | +
|
| 97 | + with it.randomize_with(solve_fail_debug=1): |
| 98 | + it.a == 2 |
| 99 | +
|
| 100 | +
|
| 101 | +In the code-block above, the `srcinfo` parameter to the |
| 102 | +`randobj` decorator causes source information to be |
| 103 | +collected for constraints in the class. The solve-fail |
| 104 | +diagnostics will now be of the following form: |
| 105 | + |
| 106 | +.. code-block:: |
| 107 | +
|
| 108 | + Problem Set: 2 constraints |
| 109 | + /project/fun/pyvsc/pyvsc-partsel-rand/ve/unit/test_solve_failure.py:30: |
| 110 | + (a == 1); |
| 111 | + /project/fun/pyvsc/pyvsc-partsel-rand/ve/unit/test_solve_failure.py:38: |
| 112 | + (a == 2); |
| 113 | + |
| 114 | +
|
| 115 | +Source-information capture may also be enabled globally |
| 116 | +via an environment variable. Set VSC_CAPTURE_SRCINFO=1 to cause |
| 117 | +all source information for all random classes to be captured. |
| 118 | + |
0 commit comments