Share GDSII reader logic and simplify review output

Share record framing, name decoding, payload validation, array reference
decoding, and Arrow cell-header parsing. Accept trailing NUL padding in
names and retain located diagnostics for malformed records.

Summarize text diffs by cell, layer, and reference target. Show comparison
and omission counts before detail. Compact canonical dumps while
preserving exact fields and duplicate counts for Git textconv. Simplify
hierarchy output and retain per-object evidence in JSON.
This commit is contained in:
Jan Petykiewicz 2026-11-17 21:49:26 -08:00
commit 15e8a3b7b5

View file

@ -2,12 +2,14 @@ mod common;
use arrow_array::ffi::{FFI_ArrowArray, FFI_ArrowSchema};
use pheon::{
basic::ErrType,
basic::{ErrType, encode_real8},
last_error_message,
library::{read_library, read_selected_library, scan_library},
read_bytes, read_cells_bytes,
record::RecordHeader,
records, scan_bytes,
records,
review::{Comparison, Document, dump_records},
scan_bytes,
};
use std::ffi::CStr;
@ -126,6 +128,136 @@ fn invalid_utf8_returns_located_parse_errors_in_each_reader() {
}
}
#[test]
fn extra_name_padding_is_ignored_by_comparison_and_arrow_readers() {
let (mut bytes, (start, end)) = fixture();
let target = payload_range(&bytes, records::RTAG_SNAME);
bytes[target].copy_from_slice(b"leaf");
bytes.truncate(end as usize);
common::record(&mut bytes, records::RTAG_BGNSTR, &[0; 24]);
common::record(&mut bytes, records::RTAG_STRNAME, b"leaf");
common::record(&mut bytes, records::RTAG_ENDSTR, &[]);
common::record(&mut bytes, records::RTAG_ENDLIB, &[]);
let mut padded = bytes.clone();
for tag in [
records::RTAG_LIBNAME,
records::RTAG_STRNAME,
records::RTAG_SNAME,
] {
let payload = payload_range(&padded, tag);
let mut data = padded[payload.clone()].to_vec();
data.extend_from_slice(&[0; 2]);
let mut record = Vec::new();
common::record(&mut record, tag, &data);
padded.splice(payload.start - 4..payload.end, record);
}
let baseline = Document::parse(&bytes).unwrap();
let document = Document::parse(&padded).unwrap();
assert_eq!(document.cells["töp"].objects.len(), 2);
assert!(
Comparison::between(&baseline, &document)
.outcome()
.is_equal()
);
let full = read_library(&padded).unwrap().1;
let scan = scan_library(&padded).unwrap().1;
let ranges = [(start + 2, end + 6), (end + 6, padded.len() as u64 - 4)];
let selected = read_selected_library(&padded, &ranges).unwrap().1;
let expected = read_library(&bytes).unwrap().1;
for actual in [&full, &scan, &selected] {
for field in ["lib_name", "cell_names"] {
assert_eq!(actual.column_by_name(field), expected.column_by_name(field));
}
}
assert_eq!(&full, &selected);
}
#[test]
fn interior_nuls_in_names_are_rejected_by_comparison_and_arrow_readers() {
let (bytes, range) = fixture();
for tag in [
records::RTAG_LIBNAME,
records::RTAG_STRNAME,
records::RTAG_SNAME,
] {
let mut malformed = bytes.clone();
let payload = payload_range(&malformed, tag);
malformed[payload.start] = 0;
let offset = if tag == records::RTAG_LIBNAME {
payload_range(&malformed, records::RTAG_UNITS).end
} else {
payload.end
};
let message = "name contains an embedded NUL";
check_read_errors(&malformed, range, offset, message, true);
assert_eq!(
Document::parse(&malformed).unwrap_err(),
format!("at byte {}: {message}", payload.start - 4)
);
}
}
#[test]
fn malformed_record_frames_are_rejected_at_the_same_offset() {
let (bytes, range) = fixture();
// STRING also exercises framing in the Arrow scan's geometry skip path.
for tag in [
records::RTAG_STRNAME,
records::RTAG_XY,
records::RTAG_STRING,
] {
let start = payload_range(&bytes, tag).start - 4;
for (size, message) in [
(2_u16, "Record size 2 is smaller than its header"),
(5, "odd record length"),
(u16::MAX - 1, "truncated record"),
] {
let mut malformed = bytes.clone();
malformed[start..start + 2].copy_from_slice(&size.to_be_bytes());
check_read_errors(&malformed, range, start, message, true);
assert_eq!(
dump_records(&malformed).unwrap_err(),
format!("at byte {start}: {message}")
);
}
}
}
#[test]
fn nonpositive_unit_scales_are_rejected_by_comparison_and_arrow_readers() {
let (bytes, range) = fixture();
let units = payload_range(&bytes, records::RTAG_UNITS);
let message = "UNITS must be positive finite values";
for field in [units.start, units.start + 8] {
for scale in [0.0, -1.0] {
let mut malformed = bytes.clone();
malformed[field..field + 8]
.copy_from_slice(&encode_real8(scale).unwrap().to_be_bytes());
check_read_errors(&malformed, range, units.start, message, true);
assert_eq!(
Document::parse(&malformed).unwrap_err(),
format!("at byte {}: {message}", units.start - 4)
);
}
}
}
#[test]
fn coordinate_payloads_require_whole_pairs() {
let (mut bytes, (start, end)) = fixture();
let after_text = payload_range(&bytes, records::RTAG_TEXT).end;
let xy = payload_range(&bytes[after_text..], records::RTAG_XY);
let payload_start = after_text + xy.start;
let mut replacement = Vec::new();
common::record(&mut replacement, records::RTAG_XY, &[0; 12]);
bytes.splice(payload_start - 4..after_text + xy.end, replacement);
let message = "XY payload must contain whole coordinate pairs";
check_read_errors(&bytes, (start, end + 4), payload_start, message, false);
assert!(Document::parse(&bytes).unwrap_err().contains(message));
}
#[test]
fn large_invalid_utf8_strings_do_not_dump_their_contents() {
let (mut bytes, range) = fixture();