dcdpr / jp / 24172309547

67%

push

github

fix(storage): Retry when mid-persist dirs detected (#525)

When another process holds the atomic-swap lock, only a `.old-*` or
`.staging-*` directory exists for the conversation being saved. Any
concurrent directory scan would miss that conversation entirely, causing
it to temporarily vanish from `load_all_conversation_ids` and
`find_conversation_dir_path`.

Both functions now detect these in-flight directories and retry up to 10
× 1ms, giving the atomic rename time to complete. An orphaned in-flight
directory (where no rename ever arrives) is safely skipped.

Two new helpers support this: `load_inflight_conversation_id` extracts a
`ConversationId` from `.old-*`/`.staging-*` names, and
`scan_conversation_ids` encapsulates the retry logic extracted from
`load_all_conversation_ids`.

---------

Signed-off-by: Jean Mertz <git@jeanmertz.com>

48 of 67 new or added lines in 2 files covered. (71.64%)

20205 of 31373 relevant lines covered (64.4%)

177.6 hits per line