tests are failing for storing to two attributes with the same name

[15r10nk]

I'm currently working on #64 and have some problems with the tests for 3.7 (reproducible on master)

The pipeline which is failing is the following:
https://github.com/alexmojaki/executing/actions/runs/4362346591/jobs/7627135034

I minimized the problem down to the following line (file was stored in tests/small_samples):

__arg.__deprecated__ = __new__.__deprecated__ = __msg

part of the test output (python 3.7.15):

mapping failed
Expected one value, found 2
search bytecode Instruction(opname='STORE_ATTR', opcode=95, arg=2, argval='__deprecated__', argrepr='__deprecated__', offset=6, starts_line=None, is_jump_target=False)

I don't know if this issue could be fixed in the SentinelNodeFinder or if it should be handled as an known issue and be ignored.

I also don't know why it started to appear now.

[alexmojaki]

It's a known issue that should be ignored.

executing/tests/test_main.py

Lines 837 to 841 in 0e5096a

	if isinstance(ctx, ast.Store):
	# Assignment to attributes and subscripts is less magical
	# but can also fail fairly easily, so we can't guarantee
	# that every node can be identified with some instruction.
	continue

[15r10nk]

Interesting. I will try to figure out why it is not. Thank you.

[15r10nk]

the error is already handled here and raised in the last line.

executing/tests/test_main.py

Lines 1246 to 1303 in 0e5096a

	print("mapping failed")
	print(e)
	if isinstance(e, NotOneValueFound):
	for value in e.values:
	print(
	"value:", ast_dump(value, indent=4, include_attributes=True)
	)
	print("search bytecode", inst)
	print("in file", source.filename)
	if py11:
	print("at position", inst.positions)
	with open(source.filename) as sourcefile:
	source_code = sourcefile.read()
	dump_source(
	source_code, start_position(inst), end_position(inst)
	)
	options = []
	for node in ast.walk(ast.parse(source_code)):
	if not hasattr(node, "lineno"):
	continue
	# if {start_position(node), end_position(node)} & {start_position(inst), end_position(inst)}:
	if start_position(node) <= end_position(
	inst
	) and start_position(inst) <= end_position(node):
	options.append(node)
	for node in sorted(
	options,
	key=lambda node: (
	node.end_lineno - node.lineno,
	node.end_col_offset - node.col_offset,
	),
	):
	if (node.end_lineno - node.lineno) > (
	inst.positions.end_lineno - inst.positions.lineno
	) * 4:
	# skip nodes which are way to long
	continue
	print()
	print(
	"possible node",
	node.lineno,
	node.end_lineno,
	node.col_offset,
	node.end_col_offset,
	ast.dump(node),
	)
	raise

The code that you mentioned is never executed in this case.

[15r10nk]

I think the problem are the open issues in the SentinelNodeFinder.

executing/tests/test_main.py

Lines 679 to 685 in 0e5096a

	skip_sentinel = [
	"load_deref",
	"4851dc1b626a95e97dbe0c53f96099d165b755dd1bd552c6ca771f7bca6d30f5",
	"508ccd0dcac13ecee6f0cea939b73ba5319c780ddbb6c496be96fe5614871d4a",
	"fc6eb521024986baa84af2634f638e40af090be4aa70ab3c22f3d022e8068228",
	"42a37b8a823eb2e510b967332661afd679c82c60b7177b992a47c16d81117c8a",
	]

This issue here is just one more. I skipped test_module_files in #64 for python < 3.8 for now.

This issues should be fixed later or categorized as KnownIssues.

But don't know if I will have time for this.