A 429 retry_after was slept immediately, including on the final attempt
that goes on to exhaust max_retries — a pointless wait right before
giving up. The wait is now carried in a per-call mutable cell and slept
at the start of the next attempt instead, so it's honored before every
attempt that actually runs and never after the last one.
errors.py's docstring claimed the exported error types cover the
missing-[discord]-extra raise path; that raise is a plain RuntimeError,
not one of these types. Reworded to match.
Compressed essay-length docstrings/comments across sender.py; no
behavior change. Verified against the aioproxies twin: aiowebhooks'
proxy-key normalization already routes zero-padded ports through
urlsplit().port (parses to int, no zero-pad on render), so host:080
and host:80 already collapse to one canonical key — no code change
needed there.
Signed-off-by: disqualifier <dev@disqualifier.me>
A server-controlled 429 retry_after/Retry-After was slept verbatim with no
finiteness check or ceiling, so an inf or huge value (adversarial or a
ms-vs-s unit mismatch) could stall send() for hours outside max_retries
accounting. Non-finite values are now rejected and finite values clamped
to MAX_RETRY_AFTER (300s).
Connection/timeout errors with no proxy provider set failed one-shot,
contradicting the README's documented "normal retry" behavior and skipping
the single most retry-worthy failure class. They now raise the internal
_Retryable signal so commons.aretry retries them under max_retries, same
as a 5xx, while still returning a failed WebhookResult (never raising)
once retries are exhausted.
Signed-off-by: disqualifier <dev@disqualifier.me>
M-2: _attempt caught only (aiohttp.ClientError, asyncio.TimeoutError); an unexpected
error escaping the attempt (closed injected session -> RuntimeError, malformed proxy url
-> ValueError) propagated out of send(), breaking the documented 'never raises on a send
failure' contract. add an outer catch-all in _send_loop converting any such exception to a
falsy WebhookResult(ok=False), logged at warning with exc_info.
aiowebhooks-F3: README 429 section + changelog were stale vs the v0.1.3 'retry any 429'
fix; added the no-parseable-wait-still-retries wording and v0.1.3/v0.1.4 changelog entries.
verified by execution: closed-session (RuntimeError) and bad-proxy (ValueError) controls
both fire and now return ok=False instead of raising.
Signed-off-by: disqualifier <dev@disqualifier.me>
treat every status==429 as retryable: sleep only when retry_after parses, but raise
_Retryable either way so aretry's backoff + max_retries cap engages. previously a 429
with no body retry_after and no Retry-After header (edge/Cloudflare/generic webhook)
returned a terminal ok=False with no retry, contradicting the documented retry-on-429.
Signed-off-by: disqualifier <dev@disqualifier.me>
the attempt count used instance state (self._attempt_no), reset in _send_loop and incremented in _attempt; two concurrent send() calls on one Webhook interleaved, corrupting each other's WebhookResult.attempts. it is now a per-call mutable cell created in _send_loop and threaded into _attempt. verified under load: 400 concurrent sends, zero corrupted counts.
Signed-off-by: disqualifier <dev@disqualifier.me>
a portless proxy url produced host::user:pass / host: (extra colons), breaking the identity match against the provider pool. the port colon is now omitted when there is no port, mirroring aioproxies' canonical key.
Signed-off-by: disqualifier <dev@disqualifier.me>
the clock= constructor param was stored (self._clock) but never read — the 429
retry_after wait uses asyncio.sleep directly. it was dead code, and the CLAUDE.md
wrongly claimed it made 429 timing test-controllable. remove the param + the unused
time import, and correct the doc (tests patch commons.retry's sleep + sender.asyncio
.sleep, not a clock seam). bump the commons pin to v0.2.1 (retry attempts floor).
verified: clock param gone, constructs fine, 18/18 fix harness intact.
Signed-off-by: disqualifier <dev@disqualifier.me>
dsql