werbeluchs

Two regressions in 2.13.0 (Avada) — one real bug, one bad default

Very thanks for the quick fix, the F12 fields are gone from the notification email. Two issues surfaced afterwards; both cost me several hours to isolate, so I’m reporting them together.

1. Gibberish validator blocks against its own documented threshold
A legitimate submission (one message field, a normal German sentence) was rejected. The block log says:

protection: gibberish-validator   verdict: scored   reason_code: GIBBERISH_CONTENT"1 of 1 text fields and 1 of 1 words scored as machine-generated (thresholds: 3 fields or 5 words, mode monitor)"

The threshold is 3 fields / 5 words, the observed values are 1 and 1 — below the threshold, yet the verdict is block. The mode is reported as monitor, which should not block at all. Additionally the word count is wrong: the field contained a full sentence, counted as 1 word.
The module is new in 2.13.0 and enables itself by default on update (protection_gibberish_enable was unset, defaulting to 1). Every Avada user gets this after updating.

2. IP protection rejects without writing a block log entry
With protection_ip_max_retries = 3 / 300 s and a 1 h block, testing the form three times locks out your own IP. That part is expected — but the rejection produces no entry in the block log, even with protection logging enabled. From the outside it looks like a silent failure: the form shows a generic error, nothing is logged, and there is no way to tell which module rejected. I had to bisect all nine protection modules to find it. A log entry with reason_code = IP_BLOCKED would have made this a one-minute diagnosis.

Environment: SilentShield 2.13.0 · Avada 7.16 (Fusion Builder Forms) · WordPress, PHP 8.4

Workaround here: gibberish validator disabled, IP thresholds relaxed to 10 retries / 15 s / 900 s. Everything else runs fine and the form spam is still fully blocked.