[rob-lake]

Fixes #99.

erase() built page boundaries with range(start, end, erase_size) and erased each consecutive pair. range excludes end, so the final page (end - erase_size, end) was never erased, although flash() still writes to it.

Changes:

• Iterate over page start addresses and erase (page_start, page_start + erase_size) for each one, so all N pages in [start, end) are erased. total_pages is now N, so the progress bar is correct.
• The BadAddress workaround still erases (page_start - erase_size, end), the same range as before.
• Raise ValueError up front if the range is not a multiple of erase_size, the same check erase_flash makes. Before, a trailing partial page was silently skipped.
• Remove the pairwise backport, which is no longer used.

Tests:

• New unit tests for erase() (with erase_flash and print_progress monkeypatched): all pages are erased including the last, the number of erase commands equals the number of pages, progress ends at 100 %, a misaligned range is rejected, and the BadAddress workaround still covers the range through end. All of them fail on current main.
• The three test_cli[...] recordings now include the extra ERASE_FLASH command for page 0x2A000 and its SUCCESS response. I don't have the test device, so I spliced these in by hand rather than re-recording them. The bytes follow the recorded response to the previous page's erase, but you may want to re-record with pytest --record -k test_cli.

Local results: pytest 23 passed, and black, isort and mypy --strict are clean. ruff check . shows 7 findings with ruff 0.17.1 (CPY001 on every module and RUF036 in main), and current main shows the same 7.

Note: tox -e test runs pytest --replay, but pytest-reserial 0.6 dropped --replay (replay is now the default), so that command fails with a fresh install. I ran coverage run -m pytest instead and left tox.ini alone.

🤖 Generated with Claude Code