John Preston
97bc3dd39e
Task: 2026/09/02/add-code-500-probe-space-class-check-and-na-rows-to-the-harness
63 lines
3.2 KiB
C++
63 lines
3.2 KiB
C++
/*
|
|
This file is part of Telegram Desktop,
|
|
the official desktop application for the Telegram messaging service.
|
|
|
|
For license and copyright information please follow this link:
|
|
https://github.com/telegramdesktop/tdesktop/blob/master/LEGAL
|
|
*/
|
|
#pragma once
|
|
|
|
#include "base/basic_types.h"
|
|
|
|
namespace Test {
|
|
|
|
class Probe;
|
|
class Runner;
|
|
|
|
// Records one "rpc retry code=<code> type=<type> request=<constructor>" row
|
|
// for an MTP answer the transport auto-resends without ever calling the
|
|
// request's fail handler. |request| is the boxed body constructor id,
|
|
// printed as 0x-prefixed lowercase hex; never a request or answer byte,
|
|
// never a payload, never a secret. No-op unless Active(), so the call site
|
|
// in application code needs no condition around it, and the Release
|
|
// branch of this module defines this function alone.
|
|
void RecordRpcRetry(int code, const QString &type, uint32 request);
|
|
|
|
// The record those rows go into. Debug-only: like BlockedLaunches() in
|
|
// test_launch_fuse.h it is declared here and deliberately left undefined
|
|
// in a Release build, because no production code reads it.
|
|
[[nodiscard]] Probe &RpcRetryProbe();
|
|
|
|
// The seam's own self-test. MTP::Instance::Private::onErrorDefault resends
|
|
// every code-500 - and every negative-code - answer after a doubling delay
|
|
// and returns true, so rpcErrorOccured returns false and the request's own
|
|
// .fail() is never reached: a scenario waiting on a product row sees no
|
|
// .done(), no .fail() and no product outcome at all, only its own stage
|
|
// timeout, while the application Debug log repeats "RPC Info: error
|
|
// received, code 500, type <T>". RecordRpcRetry is what makes that visible
|
|
// in the harness log, and RpcRetryProbe() is where a scenario reads it -
|
|
// through a mark() taken before the action, never over the whole history.
|
|
//
|
|
// The self-test issues harmless, idempotent MTPhelp_GetConfig() requests
|
|
// through an ordinary MTP::Sender - with its default FailSkipPolicy::Simple
|
|
// and never handleAllErrors() or handleFloodErrors(), because a fail
|
|
// handler that returns true for a 500 makes rpcErrorOccured return before
|
|
// onErrorDefault and the seam would then never run - and delivers
|
|
// synthesized rpc_error answers for them through the already public
|
|
// MTP::Instance::processCallback. One code 500 must record exactly one row
|
|
// naming that request's own constructor id and must leave the request
|
|
// registered for the delayed resend; one code 400 must record no row and
|
|
// must reach that request's .fail(); and a third leg delivers a 500 for a
|
|
// request id this process never sent, so the "an answer that finds no
|
|
// pending request" refusal has positive evidence instead of a guard that
|
|
// never fires in a healthy run. Every leg reads its rows in the same
|
|
// main-thread turn that called processCallback(), which is this
|
|
// self-test's own reading that the call reached the seam synchronously.
|
|
//
|
|
// It needs no wallet, no fixture secret, no chats list and no funded value;
|
|
// the only thing it asks of the process is a ready session to send through.
|
|
// It appends its own teardown as its last stage, and it emits no deliberate
|
|
// failure - every stage is expected to PASS on a healthy harness.
|
|
void AppendRpcRetrySelfTest(not_null<Runner*> runner);
|
|
|
|
} // namespace Test
|