caplog.get_records and caplog.clear conflict

Description

caplog.get_records() gets decoupled from actual caplog records when caplog.clear() is called. As a result, after caplog.clear() is called, caplog.get_records() is frozen: it does not get cleared, nor does it get new records.

During test set up it is set to the same list as caplog.records, but the latter gets replaced rather than cleared in caplog.clear(), which diverges the two objects.

Reproductive example

import logging

def test(caplog) -> None:
    def verify_consistency() -> None:
        assert caplog.get_records("call") == caplog.records

    verify_consistency()
    logging.warning("test")
    verify_consistency()
    caplog.clear()
    verify_consistency()  # fails: assert [<LogRecord: ...y, 8, "test">] == []

Environment details

Arch Linux, Python 3.9.10:

Package    Version
---------- -------
attrs      21.4.0
iniconfig  1.1.1
packaging  21.3
pip        22.0.4
pluggy     1.0.0
py         1.11.0
pyparsing  3.0.8
pytest     7.1.1
setuptools 60.10.0
tomli      2.0.1
wheel      0.37.1

Fix stage caplog records not clear

I would like to fix this issue

It seems that when setup/call/teardown stage begin, will init a stage log records stash, which used in LogCaptureFixture.get_records,
but after LogCaptureFixture.clear, the stage records in stash not be reset

I tried to fix with the following changed

# _pytest/logging.py
def _runtest_for(self, item: nodes.Item, when: str) -> Generator[None, None, None]:
    """Implement the internals of the pytest_runtest_xxx() hooks."""
    with catching_logs(
            self.caplog_handler,
            level=self.log_level,
    ) as caplog_handler, catching_logs(
        self.report_handler,
        level=self.log_level,
    ) as report_handler:
        caplog_handler.reset()
        caplog_handler.set_when(when)  # Add "when" property to LogCaptureHandler, set stage in runtest
        report_handler.reset()
        item.stash[caplog_records_key][when] = caplog_handler.records
        item.stash[caplog_handler_key] = caplog_handler

        yield

        log = report_handler.stream.getvalue().strip()
        item.add_report_section(when, "log", log)

# _pytest/logging.py
def clear(self) -> None:
    """Reset the list of log records and the captured log text."""
    self.handler.reset()
    self._item.stash[caplog_records_key][self.handler.when] = self.records    # consistent with records after clear

Fix stage caplog records not clear

It's not really clear to me why you added the set_when line, but that's probably just my inexperience with this codebase. The other line you added seems to address the issue I raised, though I am wondering if it might not be simpler to just add a layer of indirection: instead of storing List[logging.LogRecord], store LogCaptureHandler, then dynamically fetch the records when requested. It would have the advantage that the state is always consistent, and nothing needs to be rebound on clear. Again, I'm really not familiar with this code base so I might be missing something.

My understanding is StashKey[Dict[str, List[logging.LogRecord]]]() is used to storage caplog record in different stage(setup/call/teardown), but when caplog.clear() in stage, LogCaptureFixture can't know which stage caplog to clear. So I added set_when to fix it.

Can you share more details about the suggestion "instead of storing List[logging.LogRecord], store LogCaptureHandler "?

I see, thanks for the explanation, that makes sense.

While typing out a clarification of what I meant with my suggestion I realized I was missing something and it makes little sense. You can safely ignore that part.

Fix stage caplog records not clear

Update root cause and fix

caplog_handler.reset() set LogCaptureHandler.records a new list object, which lost bind to item.stash

As each running stage use one caplog_handler object and it's records object reference in item.stash item.stash[caplog_records_key][when] = caplog_handler.records. So after clear, any log records can not be captured to item.stash

Use self.records.clear() instead of self.records = [] when caplog.clear() to fix it

chore(deps): update dependency pytest to v7 (#3)

Fix stage caplog records not clear

Navigation Menu

Repository navigation

caplog.get_records and caplog.clear conflict #9877

Description

Reproductive example

Environment details

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

caplog.get_records and caplog.clear conflict #9877

Description

Description

Reproductive example

Environment details

Activity

EmptyRabbit commented on Jun 15, 2022

sanderr commented on Jun 17, 2022

EmptyRabbit commented on Jun 20, 2022

sanderr commented on Jun 20, 2022

EmptyRabbit commented on Jul 5, 2022

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions