Keep prepared search readers warm and verify abrupt indexing recovery

This commit is contained in:
karl committed 2026-10-13 01:36:34 +02:00
1 parent 423a1ea57d
commit 6b78c8bb7f
17 files changed
+1011 -41

No files matched your search

+11
View File
@@ -80,6 +80,17 @@ SEARCHI_TEST_FFMPEG and SEARCHI_TEST_FFPROBE. Run the actual Google integration:
cargo test -p searchi-service --test publication_google_cpu --test incremental_google_cpu --test google_cpu --locked -- --ignored --nocapture --test-threads=1
```
Abrupt process recovery uses the same verified asset/helper variables:
```sh
cargo test -p searchi-cli --test crash_recovery --locked -- --ignored --nocapture --test-threads=1
cargo test -p searchi-service --test prepared_google_cpu --locked -- --ignored --nocapture --test-threads=1
```
These tests use actual Google CPU inference. Crash tests terminate only their
owned CLI process tree. Do not equate process termination with power-loss proof,
or tiny prepared-reader timings with large-corpus/native UI performance.
Completed-group reuse must preserve actual vector bytes and remap current source
citations. Verify exact parser-input provenance and reject stale extraction before
encoding; legacy records require a normal rescan. Do not substitute simulated vectors or treat zero encoder calls as a