[dboeckenhoff]

#3

[dboeckenhoff]

Problem

tb thread <id> consistently returns {"thread":[],"count":0} regardless of message.

Root causes

1. getFull() doesn't expose RFC 2822 headers

The thread handler extracted References/In-Reply-To/Message-ID from
messenger.messages.getFull(msgId).headers, but these keys are absent in
practice — getFull() doesn't populate the top-level message header map
reliably. The existing /messages/:id/raw endpoint proves getRaw() has
them.

Fix: parse headers from getRaw() instead, with folded-header unfolding.

2. Angle brackets not stripped before messages.query({ headerMessageId })

The References header contains IDs in <msgid@domain> form. The stored
msg.headerMessageId field omits the brackets. Querying with brackets
included returns nothing.

Fix: strip < > from each extracted token before querying.

3. Downstream messages not found

The current algorithm only finds messages referenced by the queried message
(ancestors). Replies that arrived after it don't appear in its own References
chain.

Fix: after the References pass, run a subject-based
messages.query({ subject: normalizedSubject }) to catch downstream replies.
Strips Re:/WG:/AW:/Fwd:/FW: prefixes and [list-name] brackets before
searching.

Bonus: headerMessageId not exposed in /messages/search

messenger.messages.query() supports headerMessageId but the HTTP search
endpoint didn't pass it through. One-line fix.

Tests

• extension/src/thread-utils.js: pure functions extracted from the fix
  (header parsing, bracket stripping, subject normalization) so they can be
  unit-tested outside Thunderbird's extension context.
• test/thread-utils.test.mjs: 29 unit tests. Caught two real edge-case bugs
  during test writing (missing trim before regex), fixed in both files.
• test/quick-test.mjs: 3 new assertions — headerMessageId forwarding,
  thread returns sorted array, count matches length.

[vitalio-sh]

Thank you @dboeckenhoff — your root-cause analysis (getFull() missing the headers, angle brackets in headerMessageId queries) was exactly right, and the fix ships in 1.1.0 via #25 with you as co-author. Verified against a live Thunderbird with Gmail threads.

A few adjustments made during integration:

• thread-utils.js is now loaded by the manifest and used by background.js, so the tests exercise the real code instead of a copy.
• Subject-based downstream matches must equal the normalized subject exactly (Thunderbird's subject query is substring), exclude junk, and are labelled threadMatch: "subject" vs "references".
• In-Reply-To is parsed like References, and the getFull() fallback strips brackets too.
• /recent also honours unreadOnly, which the CLI sends but the extension ignored.

Closing in favour of #25.