Fix URL detector truncating wrapped URLs by flattening buffer

Replace fragile line-by-line reassembly heuristic with a simpler
approach: flatten the buffer by converting blank lines to spaces
(URL terminators) and stripping remaining newlines (PTY wraps),
then match URLs with a single regex on the flat string.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

app/src/lib/urlDetector.ts

@@ -2,8 +2,13 @@
  * Detects long URLs that span multiple hard-wrapped lines in PTY output.
  *
  * The Linux PTY hard-wraps long lines with \r\n at the terminal column width,
- * which breaks xterm.js WebLinksAddon URL detection. This class reassembles
+ * which breaks xterm.js WebLinksAddon URL detection. This class flattens
- * those wrapped URLs and fires a callback for ones >= 100 chars.
+ * the buffer (stripping PTY wraps, converting blank lines to spaces) and
+ * matches URLs with a single regex, firing a callback for ones >= 100 chars.
+ *
+ * When a URL match extends to the end of the flattened buffer, emission is
+ * deferred (more chunks may still be arriving). A confirmation timer emits
+ * the pending URL if no further data arrives within 500 ms.
  */
 
 const ANSI_RE =
@@ -11,6 +16,7 @@ const ANSI_RE =
 
 const MAX_BUFFER = 8 * 1024; // 8 KB rolling buffer cap
 const DEBOUNCE_MS = 300;
+const CONFIRM_MS = 500; // extra wait when URL reaches end of buffer
 const MIN_URL_LENGTH = 100;
 
 export type UrlCallback = (url: string) => void;
@@ -19,7 +25,9 @@ export class UrlDetector {
   private decoder = new TextDecoder();
   private buffer = "";
   private timer: ReturnType<typeof setTimeout> | null = null;
+  private confirmTimer: ReturnType<typeof setTimeout> | null = null;
   private lastEmitted = "";
+  private pendingUrl: string | null = null;
   private callback: UrlCallback;
 
   constructor(callback: UrlCallback) {
@@ -35,8 +43,14 @@ export class UrlDetector {
       this.buffer = this.buffer.slice(-MAX_BUFFER);
     }
 
-    // Debounce — scan after 300 ms of silence
+    // Cancel pending timers — new data arrived, rescan from scratch
     if (this.timer !== null) clearTimeout(this.timer);
+    if (this.confirmTimer !== null) {
+      clearTimeout(this.confirmTimer);
+      this.confirmTimer = null;
+    }
+
+    // Debounce — scan after 300 ms of silence
     this.timer = setTimeout(() => {
       this.timer = null;
       this.scan();
@@ -44,29 +58,60 @@ export class UrlDetector {
   }
 
   private scan(): void {
+    // 1. Strip ANSI escape sequences
     const clean = this.buffer.replace(ANSI_RE, "");
-    const lines = clean.replace(/\r\n/g, "\n").replace(/\r/g, "\n").split("\n");
 
-    for (let i = 0; i < lines.length; i++) {
+    // 2. Flatten the buffer:
-      const match = lines[i].match(/https?:\/\/[^\s'"]+/);
+    //    - Blank lines (2+ consecutive line breaks) → space (real paragraph break / URL terminator)
-      if (!match) continue;
+    //    - Remaining \r and \n → removed (PTY hard-wrap artifacts)
+    const flat = clean
+      .replace(/(\r?\n){2,}/g, " ")
+      .replace(/[\r\n]/g, "");
 
-      // Start with the URL fragment found on this line
+    if (!flat) return;
-      let url = match[0];
 
-      // Concatenate subsequent continuation lines (non-empty, no spaces, no leading whitespace)
+    // 3. Match URLs on the flattened string — spans across wrapped lines naturally
-      for (let j = i + 1; j < lines.length; j++) {
+    const urlRe = /https?:\/\/[^\s'"<>\x07]+/g;
-        const next = lines[j];
+    let m: RegExpExecArray | null;
-        if (!next || next.startsWith(" ") || next.includes(" ")) break;
+
-        url += next;
+    while ((m = urlRe.exec(flat)) !== null) {
-        i = j; // skip this line in the outer loop
+      const url = m[0];
+
+      // 4. Filter by length
+      if (url.length < MIN_URL_LENGTH) continue;
+
+      // 5. If the match extends to the very end of the flattened string,
+      //    more chunks may still be arriving — defer emission.
+      if (m.index + url.length >= flat.length) {
+        this.pendingUrl = url;
+        this.confirmTimer = setTimeout(() => {
+          this.confirmTimer = null;
+          this.emitPending();
+        }, CONFIRM_MS);
+        return;
       }
 
-      if (url.length >= MIN_URL_LENGTH && url !== this.lastEmitted) {
+      // 6. URL is clearly complete (more content follows) — dedup + emit
+      this.pendingUrl = null;
+      if (url !== this.lastEmitted) {
         this.lastEmitted = url;
         this.callback(url);
       }
     }
+
+    // Scan finished without a URL at the buffer end.
+    // If we had a pending URL from a previous scan, it's now confirmed complete.
+    if (this.pendingUrl) {
+      this.emitPending();
+    }
+  }
+
+  private emitPending(): void {
+    if (this.pendingUrl && this.pendingUrl !== this.lastEmitted) {
+      this.lastEmitted = this.pendingUrl;
+      this.callback(this.pendingUrl);
+    }
+    this.pendingUrl = null;
   }
 
   dispose(): void {
@@ -74,5 +119,9 @@ export class UrlDetector {
       clearTimeout(this.timer);
       this.timer = null;
     }
+    if (this.confirmTimer !== null) {
+      clearTimeout(this.confirmTimer);
+      this.confirmTimer = null;
+    }
   }
 }