This patch adds a robust, optional mechanism for handling "headerless" input—
log messages that do not conform to traditional syslog formatting.
- **Headerless detection (opt-in)**
- Controlled by the new `detect.headerless` boolean (default: off)
- Detects messages with **no PRI** and **no valid timestamp**
- Excludes structured inputs (e.g. JSON starting with `{` or `[`) as
before
- Injects default `hostname` and `tag` values
- Flags message internally as `HEADERLESS_MSG` for further processing
- **Fallback processing options**
- `headerless.ruleset`: route headerless messages to a dedicated ruleset
- `headerless.errorfile`: optionally store raw input to a file
- `headerless.drop`: discard headerless messages early if desired
- **Thread-safe HUP signal handling**
- New `doHUPParser` entry point allows safe log rotation for error file
- Follows standard reopen-on-write pattern post-HUP
- **Testing & Maintenance**
- Adds two test cases: `pmrfc3164-headerless.sh` and `pmrfc3164-drop.sh`
- Extends documentation for all new parameters
- Cleans up code formatting, includes, and bumps copyright
Some environments produce mixed or malformed input streams. This patch enables
early, lightweight detection of non-syslog input, with customizable recovery
and routing strategies. It avoids unnecessary parsing work and gives operators
better tools to isolate or discard garbage input—without breaking legacy behavior.
