Fix base64 last decoding offset (#551)

packages/wasm-util/src/base64.ts

@@ -3,33 +3,33 @@
 
 // Use an array for our indexer - this is faster than using map access. In
 // this case we assume ASCII-only inputs, so we cannot overflow the array
-const chr = 'ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/';
+const CHR = 'ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/' as const;
 const map = new Array<number>(256);
 
 // We use charCodeAt for access here and in the decoder loop - this is faster
 // on lookups (array + numbers) and also faster than accessing the specific
 // character via data[i]
-for (let i = 0, count = chr.length; i < count; i++) {
-  map[chr.charCodeAt(i)] = i;
+for (let i = 0, count = CHR.length; i < count; i++) {
+  map[CHR.charCodeAt(i)] = i;
 }
 
 /**
  * @name base64Decode
  * @description
- * A base64Decoding function that operates in all environments. Unlike decoding
+ * A base64 decoding function that operates in all environments. Unlike decoding
  * from Buffer (Node.js only) or atob (browser-only) this implementation is
  * slightly slower, but it is platform independent.
  *
- * For our usage, since we have access to the static final size (where used), we
- * decode to a specified output buffer. This also means we have applied a number
- * of optimizations based on this - checking output position instead of chars.
+ * For our usage, since we have access to the static final size, so we decode
+ * to a specified output buffer. This also means we have applied a number of
+ * optimizations based on this - checking output position instead of chars.
  */
 export function base64Decode (data: string, out: Uint8Array): Uint8Array {
   let byte = 0;
   let bits = 0;
   let pos = -1;
 
-  for (let i = 0, count = out.length; pos < count; i++) {
+  for (let i = 0, last = out.length - 1; pos !== last; i++) {
     // each character represents 6 bits
     byte = (byte << 6) | map[data.charCodeAt(i)];