fix(app): prevent OverflowError from addresses above 2^63 in memory map

[cromachina]

Why is this PR necessary, what does it do?

Fixes an error that appears when viewing the memory map (and also potentially the module viewer). The error prevents the lists from being fully populated. The error can also potentially segfault the program, at least when shiboken stuff runs.

Also includes a minor cosmetic change to make comparing the addresses in a list easier (monospace font for the addresses).

Checklist (complete all items):

• [✅] Added tests as necessary.
• [✅] There is no breaking change for existing features.

References:

Fixes #73

Notes:

No notes to be shared.

[cromachina]

The "Lint PR title" check doesn't check case insensitive, even though github generates titles from branches automatically with casing.

[JeanExtreme002]

Hi, @cromachina! Thanks for contributing.

There are some issues that must be fixed before merging this PR. I'll post the reviews here.

[JeanExtreme002]

Good catch on the root cause. Verified locally: regular addresses (0x7ffd…) go through setData fine — only values above 2**63 overflow, which is exactly [vsyscall] at 0xffffffffff600000 on Linux x86-64. The Python-side payload is the right call, and it also covers pointer_scan_dialog.py:687, which had the same latent bug.

Three things before merge, inline. The first one is a crash, the other two are smaller.

One more: the checklist says tests were added, but there are none in the diff and _widgets.py has no test file yet. tests/app/ + QT_QPA_PLATFORM=offscreen already make this cheap to cover — a round trip with 0xffffffffff600000 and a sort over rows with a missing payload would pin both regressions.

[JeanExtreme002]

Thanks for the quick turnaround — roles, type hints and the sort recursion are all sorted (verified: sorting a column with None payloads no longer crashes, and UserRole / UserRole + 1 stay distinct). Resolved those threads.

One new issue from the text() override, plus two smaller things, inline.

For the record on the docstring: the cap isn't really about unsigned support — a plain QStandardItem takes 0x7FFDABCD1234 fine and only breaks past 2**63, i.e. QVariant's qint64. Your 2**128 case shows the same edge. Also worth stating the constraint the workaround introduces: the payload never reaches the C++ model, so it must be read as item.data(role) and never model.data(index, role) — that path converts back to a QVariant and overflows again (confirmed locally).

[JeanExtreme002]

Adding applicable suggestions for the two open points above. Both verified locally: with them the app suite is 83 passed and flake8 is clean; putting the text() override back makes the new assertions fail.

[JeanExtreme002]

I went ahead and pushed the two suggestions to this branch (a82087b) since the branch allows maintainer edits — feel free to amend or revert if you'd rather do it yourself.

• Dropped the text() override. Its except KeyError was unreachable (dict.get() returns None), so text() was returning the raw payload instead of the label, and the __lt__ fallback was comparing 'None' to 'None' instead of the labels.
• Added assertions to the NumericItem test, plus coverage for the role collision and for model.sort() over a column mixing payloads with None — the case that actually segfaulted, which comparing two items directly doesn't reach.

Verified locally: 406 passed, 78 skipped, flake8 clean.

Still open: the model.data() note in the docstring, and _MONO living in pointer_scan_dialog rather than _widgets. Both are minor.

[JeanExtreme002]

Pushed the last two as well (57bb6ad):

• MONOSPACE_FAMILY now lives in _widgets.py next to NumericItem, and all three dialogs read it from there instead of importing _MONO across dialog modules. The memory map's size field was hardcoding the same string, so it uses the constant too.
• The NumericItem tests moved to tests/app/test_app_widgets.py. They were behind a pytest-qt skip they never needed (no qtbot, just a QApplication), so they run wherever PySide6 is installed.

406 passed, 78 skipped, flake8 clean. All review threads are resolved now — the only thing I left alone is the docstring wording, which is fine as is.

[JeanExtreme002]

Last one from me (9fcfc0a) — reworded the docstring rationale. A plain QStandardItem takes 0x7FFDABCD1234 fine and only breaks past 2**63, so the cap is QVariant's qint64 rather than unsigned support (your 2**128 case is the same edge). Also noted the constraint the workaround creates: the payload never reaches the C++ model, so reading it back via model.data(index, role) hits the same OverflowError — worth having written down before someone adds a delegate or a sort role over these columns.

From my side this is ready to merge.

[JeanExtreme002]

LGTM 🚀