0.60.0rc1 regression: NameError: name 'length_of_iterator' is not defined

[sk1p]

Reporting a bug

• I have tried using the latest released version of Numba (most recent is
  visible in the release notes
  (https://numba.readthedocs.io/en/stable/release-notes-overview.html).
• I have included a self contained code sample to reproduce the problem.
  i.e. it's possible to run as 'python bug.py'.

The following minimal example fails to compile with 0.60.0rc1 (works fine with 0.59.1):

import numpy as np
import numba


@numba.njit
def _inner():
    range_start = 0
    for _ in range(1):
        np.array([
            1 for _ in range(range_start, 7)
        ])
        range_start = 0

_inner()

Error message:

../numba/numba/core/dispatcher.py:423: in _compile_for_args
    error_rewrite(e, 'typing')
../numba/numba/core/dispatcher.py:362: in error_rewrite
    raise e
../numba/numba/core/dispatcher.py:375: in _compile_for_args
    return_val = self.compile(tuple(argtypes))
../numba/numba/core/dispatcher.py:905: in compile
    cres = self._compiler.compile(args, return_type)
../numba/numba/core/dispatcher.py:84: in compile
    raise retval
../numba/numba/core/dispatcher.py:94: in _compile_cached
    retval = self._compile_core(args, return_type)
../numba/numba/core/dispatcher.py:107: in _compile_core
    cres = compiler.compile_extra(self.targetdescr.typing_context,
../numba/numba/core/compiler.py:744: in compile_extra
    return pipeline.compile_extra(func)
../numba/numba/core/compiler.py:438: in compile_extra
    return self._compile_bytecode()
../numba/numba/core/compiler.py:506: in _compile_bytecode
    return self._compile_core()
../numba/numba/core/compiler.py:485: in _compile_core
    raise e
../numba/numba/core/compiler.py:472: in _compile_core
    pm.run(self.state)
../numba/numba/core/compiler_machinery.py:368: in run
    raise patched_exception
../numba/numba/core/compiler_machinery.py:356: in run
    self._runPass(idx, pass_inst, state)
../numba/numba/core/compiler_lock.py:35: in _acquire_compile_lock
    return func(*args, **kwargs)
../numba/numba/core/compiler_machinery.py:311: in _runPass
    mutated |= check(pss.run_pass, internal_state)
../numba/numba/core/compiler_machinery.py:273: in check
    mangled = func(compiler_state)
../numba/numba/core/typed_passes.py:112: in run_pass
    typemap, return_type, calltypes, errs = type_inference_stage(
../numba/numba/core/typed_passes.py:91: in type_inference_stage
    infer.build_constraint()
../numba/numba/core/typeinfer.py:1044: in build_constraint
    self.constrain_statement(inst)
../numba/numba/core/typeinfer.py:1406: in constrain_statement
    self.typeof_assign(inst)
../numba/numba/core/typeinfer.py:1481: in typeof_assign
    self.typeof_global(inst, inst.target, value)
../numba/numba/core/typeinfer.py:1581: in typeof_global
    typ = self.resolve_value_type(inst, gvar.value)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

self = <numba.core.typeinfer.TypeInferer object at 0x7fb972830bc0>, inst = <numba.core.ir.Assign object at 0x7fb9727f3740>, val = <intrinsic length_of_iterator>

    def resolve_value_type(self, inst, val):
        """
        Resolve the type of a simple Python value, such as can be
        represented by literals.
        """
        try:
            return self.context.resolve_value_type(val)
        except ValueError as e:
            msg = str(e)
>       raise TypingError(msg, loc=inst.loc)
E       numba.core.errors.TypingError: Failed in nopython mode pipeline (step: nopython frontend)
E       NameError: name 'length_of_iterator' is not defined

../numba/numba/core/typeinfer.py:1502: TypingError

This seems to be some kind of "import order" problem - adding import numba.core.inline_closurecall before the _inner() call fixes the issue. Patching enable_inline_arraycall to False also fixes the issue. Thanks!

[gmarkall]

Thanks for the report and reproducer! I can confirm that this reproduces with 0.60.0rc1, but not with 0.59.1.

[gmarkall]

The commit that caused this is: d184335 from #9437 (cc @kc611).

That commit moved length_of_iterator into inline_closurecall, so it is no longer registered when the CPU target is set up (it was moved from numba.cpython.rangeobj.py, which is imported in CPUContext.load_additional_registries() - this explains why importing inline_closurecall seems to resolve the issue.