edfcore 0.4.319 → 0.4.321

package/dist/constants.d.ts

@@ -109,5 +109,5 @@ export declare const SIGNAL_FIELD_BLOCK_OFFSETS: {
     readonly reserved: 224;
 };
 /** Published package version. Kept in sync with package.json by a test. */
-export declare const VERSION = "0.4.319";
+export declare const VERSION = "0.4.321";
 //# sourceMappingURL=constants.d.ts.map

package/dist/constants.js

@@ -79,5 +79,5 @@ export const SIGNAL_FIELD_BLOCK_OFFSETS = {
     reserved: 224,
 };
 /** Published package version. Kept in sync with package.json by a test. */
-export const VERSION = '0.4.319';
+export const VERSION = '0.4.321';
 //# sourceMappingURL=constants.js.map

package/docs/CHANGELOG.md

@@ -6,6 +6,34 @@ alone does not tell you whether you were affected.
 edfcore is pre-1.0. Patch releases have carried behaviour changes where the old behaviour was a
 defect; those are called out below.
 
+## 0.4.321
+
+- **Added** a guard for the offset rule `edf-format.md` states: offsets stay in plain floats, which
+  are exact to 2^53, because a data offset in a multi-gigabyte BDF crosses 2^31 and a bitwise
+  operator wraps it negative there without warning. Nothing checked it, and the failure is the
+  silent kind — a 22-hour BDF passes 2^31 bytes about nine hours in, so a truncated offset would
+  read plausible samples from the wrong place for the whole back half of the recording.
+- Not a ban: `decode/digital.ts` assembles every sample with `|` and `<<` and must, and the same
+  page prints those two lines as the definition of the format. The guard is an inventory instead —
+  every bitwise operator in `src/` has to sit in one of three modules with a written reason, which
+  makes adding a fourth a deliberate act. `record-index.ts` is one of the three, and its entry
+  records why `(low + high) >> 1` is safe: the record count comes from an eight-character header
+  field, so the index cannot approach 2^30.
+- The demonstration corrects a detail the page leaves implicit. `| 0` and `>>` wrap the offset
+  negative at 2^31, but `>>>` is unsigned and returns something plausible until 2^32 — about
+  eighteen hours into the same recording — which is the more dangerous of the two behaviours.
+
+## 0.4.320
+
+- **Extended** 0.4.319 to the second copy of the harness table. `scripts/golden/README.md` carries
+  the complete one — four rows rather than the documentation page's three, because it includes
+  `corpus-parity.test.ts`, the only harness whose inputs nobody here chose — and `tests/README.md`
+  sends the reader to it for "what each harness claims and how strong that claim is".
+- That copy names each harness by file, which is the stronger form, so a row cannot outlive the
+  test it describes. Every named file has to exist, every bit-for-bit row has to belong to a
+  harness that compares with `Object.is` and no tolerance, and the MNE bound has to match both the
+  constant in the harness and the figure the other table publishes.
+
 ## 0.4.319
 
 - **Added** a check that the three cross-implementation harnesses claim on the page exactly what

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "edfcore",
-  "version": "0.4.319",
+  "version": "0.4.321",
   "description": "Modern, typed, zero-dependency reader for EDF, EDF+, BDF and BDF+ biosignal files. Works in browsers and Node with true random access.",
   "keywords": [
     "edf",

package/src/constants.ts

@@ -93,4 +93,4 @@ export const SIGNAL_FIELD_BLOCK_OFFSETS = {
 } as const;
 
 /** Published package version. Kept in sync with package.json by a test. */
-export const VERSION = '0.4.319';
+export const VERSION = '0.4.321';