Fix VIRTNBD_FALSE_SUCCESS false-positive on Windows VMs with BitLocker

[hostarts]

Summary

perform_backup()'s post-backup ERROR-line scan used a case-insensitive substring match (grep -qi "ERROR"), which fires on any occurrence of the word "ERROR" in the captured virtnbdbackup log — including INFO/WARN payloads. On Windows VMs the QEMU guest agent surfaces BitLocker status text that legitimately contains the word "ERROR" during normal operation, so otherwise-successful backups are flagged as VIRTNBD_FALSE_SUCCESS, retried, and generate spurious alert noise.

The check is now anchored to the timestamped severity prefix:

^[YYYY-MM-DD HH:MM:SS] ERROR 

so only real ERROR-level log records trigger the guard. ANSI colour escapes emitted by virtnbdbackup are stripped before the anchor match so coloured output still matches. The same fix is applied to the diagnostic-snippet extraction that follows.

What changed

• vmbackup.sh (perform_backup, ~line 3590): switched the false-success guard and the snippet extractor to sed -E 's/\x1b\[[0-9;]*m//g' | grep -qE '^\[[0-9-]{10} [0-9:]{8}\] ERROR '.
• CHANGELOG.md: added [Unreleased] entry under ### Fixed.

Test plan

• In production, the failing run was a Windows VM whose guest-agent output contained the substring ERROR inside a BitLocker status line; old grep matched, new grep does not.
• Verified anchored regex still catches real [2026-05-01 12:34:56] ERROR ... lines emitted by virtnbdbackup.
• bash -n vmbackup.sh passes.

🤖 Generated with Claude Code