[dormancygrace]

Log snapshots always ran an ANSI regexp and classified every rune, even for ordinary ASCII text. Skip the regexp when no ESC is present and bypass Unicode classification for printable ASCII, LF and TAB, retaining secret redaction and fallback behavior.

The committed three-process, 200 ms sweep on the 1 GHz C906 measured a median 478.435 ms -> 333.671 ms (-30.26%) for the existing ~1 MiB ASCII DHCP-log fixture. Allocated bytes fell by about 2 MiB per operation. Complete samples and reproduction commands are in tests/manual/c906-log-sanitize/. These are sanitizer microbenchmarks, not boot-time measurements; fallback-case differences are not claimed as separate gains.

Validation: host logs/logger race tests and vet pass, including valid UTF-8 C1 controls and secrets revealed by filtering. The implementation previously passed the device logs suite and production RISC-V cross-build; this follow-up changes tests and documentation only. Existing cases cover every input byte, invalid UTF-8, ANSI/OSC and Unicode controls. RVV work is separate.

[github-actions]

Build outputs

Compared with the build of main at c93fb1a.

1 changed, 0 added, 0 removed of 570 outputs.

	directory	files
changed	server	server/NanoKVM-Server
Hashes: outputs-sha256 artifact of this run.

[dormancygrace]

The change is correct. It keeps every guarantee of the original filter:

• ESC skip: the regexp can only match text that starts with ESC, so skipping it when no ESC is present is exact.
• Rune-filter bypass: only printable ASCII, LF and TAB skip the rune filter. C0 controls, DEL, CR, NUL, ESC and all non-ASCII (C1 controls included) still reach it.
• Redaction order: redaction still runs after filtering.

I also compared the original and the new sanitiser on 2M random inputs built from ESC and partial escape sequences, OSC, DEL, NUL, CR, C1 bytes, invalid UTF-8 and ZWSP. There were no differences. go test -tags teststub ./logs passes.

Only two small comments follow.

[dormancygrace]

Reviewed the sanitization fast paths and secret-keyword filtering. The runtime behavior is preserved: ASCII text avoids redundant regexp/UTF-8 passes while escape removal, Unicode controls and redaction remain covered. Added C1-control cases, including a control embedded in a sensitive keyword, and aligned performance claims with the committed measurement artifact. Race tests and vet pass for logs/logger; the integrated merge passes the same race tests. All 15 checks on e2a16c1 pass, including firmware. Remaining findings were documentation/test coverage only; both discussions are resolved.