fix: ensure TextEncoder never returns pooled arrays

Previous impl always ensured 0-offset arrays, but not that there
is no data _past_ the returned view.

This only affects Node.js-like platforms, as only there native
buffers are pooled.

Author: ChALkeR

fallback/encoding.js

@@ -211,7 +211,8 @@ export class TextEncoder {
   encode(str = '') {
     if (typeof str !== 'string') str = `${str}`
     const res = utf8fromStringLoose(str)
-    return res.byteOffset === 0 ? res : res.slice(0) // Ensure 0-offset, to match new Uint8Array (per spec), which is non-pooled
+    // match new Uint8Array (per spec), which is non-pooled
+    return res.byteOffset === 0 && res.length === res.buffer.byteLength ? res : res.slice(0)
   }
 
   encodeInto(str, target) {

tests/encoding/nopool.test.js

@@ -0,0 +1,21 @@
+import { TextEncoder } from '@exodus/bytes/encoding.js'
+import { test } from 'node:test'
+
+test('TextEncoder returns non-pooled buffers', (t) => {
+  const encoder = new TextEncoder()
+  for (let i = 0; i < 256; i++) {
+    t.assert.strictEqual(encoder.encode('x'.repeat(128)).buffer.byteLength, 128)
+  }
+
+  for (let i = 0; i < 256; i++) {
+    t.assert.strictEqual(encoder.encode('x'.repeat(64)).buffer.byteLength, 64)
+  }
+
+  for (let i = 0; i < 512; i++) {
+    t.assert.strictEqual(encoder.encode('x'.repeat(32)).buffer.byteLength, 32)
+  }
+
+  for (let i = 0; i < 512; i++) {
+    t.assert.strictEqual(encoder.encode('').buffer.byteLength, 0)
+  }
+})