ilml-plugin-linkedin 1.20.8 → 1.20.9

package/CHANGELOG.md

@@ -1,5 +1,8 @@
 # Changelog
 
+## 1.20.9
+- **Commands now recover on their own when LinkedIn puts up its sign-in wall mid-run — instead of stopping and making you start over.** LinkedIn periodically throws its "join / sign in" screen on a page even though your session is still good (common when you've recently signed in from another device or location). Until now that ended the command with "session expired", and you had to log in again and re-run it. Now, when a command hits that wall, it quietly reopens your feed — where LinkedIn usually recognises you from your existing session within a few seconds, no password needed — refreshes your saved sign-in, and continues where it left off, hands-free. If LinkedIn genuinely asks for your email and password (a real sign-out), the command pauses and waits while you sign in **in the browser window it already has open**, then resumes on its own — no need to close anything or run the command again. Safeguards: your saved session is only overwritten when the refreshed one is actually valid, so a half-loaded page can never replace a good login; and an unattended/scheduled run won't hang waiting for a sign-in nobody can complete — it fails fast with a clear message instead. This works the same way across every command. No schema change.
+
 ## 1.20.8
 - **Fixed: `messages --push-drafts` no longer leaves a duplicate copy of the sent message in your local conversation history.** When a draft was sent, the plugin could record the message twice — once as the draft-turned-sent and once from re-reading the thread afterwards — because the two copies differed only by the paragraph breaks (`\n\n`) that LinkedIn strips when it re-extracts a thread, so they weren't recognised as the same message. Two fixes: (1) message de-duplication is now whitespace-insensitive, so any two messages that differ only in spacing/line-breaks collapse into one; (2) the send flow now writes your conversations file through a single owner instead of two competing writers that could overwrite each other. If extraction ever misses the just-sent message on a very long thread, the draft is kept (marked sent) rather than dropped, so a sent message is never lost from the record. Recognising your own sent message now tolerates the small differences LinkedIn introduces in the displayed sender name (case, spacing, an appended credential), so it can't accidentally leave a second copy. Message de-duplication is also now robust to unexpected non-text values in stored data — a single corrupted entry can no longer abort a whole sync. Existing duplicates tidy themselves up the next time that thread is synced. No schema change — your messages, contacts and drafts are otherwise untouched.