tag-txt-import Check

txt_import_check.py runs tag-txt-import on every fixture in txt-import/ and asserts what its conversion report must say: the variant, the records written, each repair, and that the round trip back to text passed. It also checks the written log's row counts and provenance; that exports run together fail; that variant A matches an independent implementation of btviz's decode; and that non-logs are refused. Run it by hand after building the tool:

python3 host/commandline/test/txt_import_check.py <build-dir>/Release/bin/tag-txt-import

It needs only Python 3's standard library. There is no test framework; it stops at the first failure.

The fixtures

Each is a few records trimmed from a file in the legacy collection, with the UUID (variant A: the device string) replaced by zeros and the firmware repository URL by a placeholder:

Fixture	Variant	What it exercises
c1-fivemin.txt	C	Current header, FIVEMIN, an older writer's fixed decimals
c1-fivemin-crlf.txt	C	The same with CRLF line endings
c1-duplicate-record.txt	C	A record repeated exactly: removed
c1-implausible-time.txt	C	A record dated 2005: dropped
c2-min.txt	C	MIN, ten 6-bit buckets
c3-packed-word.txt	C	The packed word on V: lines, checked against the buckets
b-fivemin-9x7.txt	B	The V5 header and the pre-2021 nine-bucket FIVEMIN decode, repaired
d-prestag.txt	D	PresTag blocks; V: and TC: (both voltages) stored as their mean
e-no-header.txt	E	No header; format from the data
e-concatenated.txt	E	Two exports run together, with orphan and unclosed records: must fail
header-only.txt	C	A header with no data
a-rtcv4.txt	A	BitTagRTCv4 dump, translated as btviz decodes it; the tag's clock had reset when read

When a fixture's expected counts change, update EXPECT in the script and say why in the commit.