fix(iq): report <error> detail the IQ parser drops, once per process

[jlucaso1]

Summary

RequestUtils::parse_iq_response reads code/text/type/backoff off the <error> node and drops the rest. Those four are exactly what WA Web's parseIqResponse keeps, so reading only them is the right default, but "we read what WA Web reads" and "the server sent nothing more" are different claims and only the first was ever checked: a bare bad-request gives no way to tell an empty error from a detailed one the parser discarded.

This adds a probe that names the unread parts and nothing else. It reports once per process, because the finding is a note to this library's maintainers rather than something a calling application can act on, and rejected IQs arrive in bursts (usync, prekey and app-state fan-outs all retry) where a per-occurrence warning would be noise at exactly the moment the log matters.

Protocol evidence

• docs/captured-js/WA/Parse/IqResponse.js builds its error result from i.code, i.text, i.type and i.backoff only, confirming the four attributes we read are the full set WA Web keeps.
• The same function takes a third argument for the rest: typeof n == "function" ? (l = n(e)) : (l = n.parseOrThrow(a)), where a is the <error> node. So WA Web does have an escape hatch for a call site that needs more, exposed as deprecatedSendIqErrorParser in docs/captured-js/WA/Deprecated/SendIq.js.
• Worth stating plainly, since it sets the scope: no call site in the capture passes that third argument, and the only wrapper exposing it is a deprecated one. So there is no evidence yet that rejected IQs carry anything past those four attributes. This probe is what would produce that evidence.

Changes

• dropped_error_detail names the attributes outside PARSED_ERROR_ATTRS, the child tags, and the kind of a raw payload. Names only, no values: an attribute value can hold a JID and the report lands at a level production enables. It reads content directly rather than children(), which answers None for a byte or string payload that is dropped just the same.
• PARSED_ERROR_ATTRS mirrors what the parser reads, with the parser commented to point at it. Reading a fifth attribute without adding it there is what would make the probe report a false positive on every response.
• The report is gated by a static AtomicBool and by log_enabled!, both checked before the scan: once the warning is out, or with warnings filtered off, every further rejected IQ costs one relaxed load. Racing callers settle it with a swap, so at most one message is emitted.
• warn_on_dropped_error_detail is #[cold] to keep it out of parse_iq_response, which is on the receive path and has its code size gated in CI.
• Drive-by: generate_request_id now uses the imported Ordering instead of the full path, which unused_qualifications requires once the import exists.

Cost

The happy path is untouched: nothing here runs for a successful IQ. On a rejected one the scan walks the error node's attributes and its content discriminant, and allocates nothing when there is nothing to report (a Filter yields a zero lower bound, so the collect does not allocate). The two Vecs are only ever materialized on the single occurrence that logs.

Not in this PR

If the probe does fire in production, the follow-up is to hand the detail to the caller rather than only to the log: log is a facade, so an application that installs no logger sees nothing, and one that does could already raise Client/Recv to debug. That means an IqError::ServerError field carrying the node, which is a public error-surface change touching every construction site, and it is not worth making speculatively before the probe shows there is anything to carry.

Validation

cargo fmt --all
cargo test -p wacore --lib          # 1335 passed
cargo clippy -p wacore --all-targets -- -D warnings

Full matrix left to CI.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Repository UI (base), Organization UI (inherited)

Review profile: ASSERTIVE

Plan: Pro Plus

Run ID: 657b9728-c9dd-4537-a8d6-e372f10bda56

📥 Commits

Reviewing files that changed from the base of the PR and between c06acb1 and fc4cfa3.

📒 Files selected for processing (1)

• wacore/src/request.rs

---

📝 Walkthrough

Summary by CodeRabbit

• Bug Fixes

  • Improved handling of server error responses to capture additional diagnostic details.
  • Correctly detects error information in attributes, child content, and text or byte payloads.
  • Improved handling of empty responses and payloads for more consistent behavior.

• Improvements

  • Added clearer, one-time warnings when responses contain unsupported or unparsed content.
  • Warning details are limited to safe, non-sensitive information.

Walkthrough

Changes

The IQ response parser detects unparsed error-node content and emits a redacted warning once per process. Tests cover parsed errors, attributes, child nodes, byte/text payloads, and empty payloads.

IQ error detail detection

Layer / File(s)	Summary
Error parsing and warning detection wacore/src/request.rs	IQ error handling extracts known attributes, inspects remaining attributes, child nodes, and nonempty payloads, then emits a warning once per process when details are dropped.
Error detail detection tests wacore/src/request.rs	Tests validate detection of unparsed attributes, child nodes, byte/text payloads, and suppression for parsed or empty errors and empty payloads.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Possibly related PRs

• oxidezap/whatsapp-rust#1100: Both changes modify IQ error handling in wacore/src/request.rs; that PR adds timeout and transport classification methods.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly and concisely describes the main change: reporting dropped IQ error details once per process.
Description check	✅ Passed	The description directly explains the parser change, probe behavior, scope, performance considerations, tests, and validation.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/iq-error-dropped-detail-probe

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[github-actions]

📦 Binary size report

Metric	main	PR	Δ
bin size (stripped)	9.95 MiB	9.95 MiB	+2.28 KiB (+0.02%) 🔺
bin .text	7.97 MiB	7.98 MiB	+2.00 KiB (+0.02%) 🔺
bin allocated (text+data+bss)	9.95 MiB	9.95 MiB	+32 B (+0.00%) 🔺
llvm-lines wacore	511,608	512,919	+1,311 (+0.26%) 🔺
llvm-lines wacore copies	16,704	16,742	+38 (+0.23%) 🔺
llvm-lines whatsapp-rust lib	727,450	727,450	0
llvm-lines whatsapp-rust lib copies	22,955	22,955	0
deps crates (Cargo.lock)	462	462	0

.text per crate

Crate	main	PR	Δ
.text whatsapp_rust	1.80 MiB	1.80 MiB	0
.text wacore	687.43 KiB	689.71 KiB	+2.28 KiB (+0.33%) 🔺
.text wacore_binary	91.42 KiB	91.42 KiB	0
.text wacore_libsignal	170.64 KiB	170.64 KiB	0
.text wacore_appstate	22.35 KiB	22.35 KiB	0
.text wacore_noise	21.79 KiB	21.79 KiB	0
.text waproto	1.74 MiB	1.74 MiB	0
.text whatsapp_rust_sqlite_storage	515.70 KiB	515.62 KiB	-83 B (-0.02%) 🔽
.text whatsapp_rust_tokio_transport	40.49 KiB	40.49 KiB	0
.text whatsapp_rust_ureq_http_client	11.83 KiB	11.83 KiB	0
.text std	984.91 KiB	984.91 KiB	0
.text other deps	1.90 MiB	1.90 MiB	-298 B (-0.01%) 🔽

Top movers (cargo-bloat attribution)

Crate	main	PR	Δ
wacore	687.43 KiB	689.71 KiB	+2.28 KiB (+0.33%)

Baseline: 9c3e96f84 (latest main run) · Head: 51ada0362 · Graphs

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@wacore/src/request.rs`:
- Around line 549-561: The test a_fully_parsed_error_reports_nothing currently
relies on warn_on_dropped_error_detail reaching its warning branch, which
depends on global log::Level::Warn state. Update the test to assert the
warning-branch behavior directly through the relevant dropped-error-detail
result or branch-specific helper, while preserving the existing assertion that
fully parsed errors return None; do not rely on process-wide logger
configuration.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Repository UI (base), Organization UI (inherited)

Review profile: ASSERTIVE

Plan: Pro Plus

Run ID: c12c3e1d-1932-4887-b897-0dbf4d4cd8e4

📥 Commits

Reviewing files that changed from the base of the PR and between 9c3e96f and c06acb1.

📒 Files selected for processing (1)

• wacore/src/request.rs