[MG_Pipe, MG_Backend, MG_Remote, MG_Test, scripts] (P7): the Magma wire fatals and the two Staged aborts go through Session::Fail

Five deaths in the split server image raised their own std::abort(). Three of them are
funnels - MagmaWireFatal (Renderer/WireFramebuffer.inc), WireDescriptorFatal
(Renderer/UniformManager.cpp) and WireBufferLegacyFatal (Renderer/WireDraw.inc) - and the
fifteen `@P7` refusals on the tree (8 in UniformManager.cpp, 4 in WireFramebuffer.inc, 2 in
WireDraw.inc, 1 in VulkanRenderer.cpp; plan section 1.3's count confirmed) all die through
two of them. Every one of those deaths published NO SessionFault frame, so the peer read a
bare EOF; bumped NO SessionFaultCount(), so exit gate S8 could neither confirm nor falsify
it; and carried a hand-typed family word with no path to FatalFamilies.def's projection.

MG_Backend may not name an MG_Remote symbol on its own account, so the route is a seam in
MG_Pipe with MGP_TRIP_WIRE_REPORT's shape (PipeApply.cpp:66, and the new code sits beside
it): MG_Pipe/PipeSessionFail.h declares MGPipeSessionFail, PipeApply.cpp defines it inside
`#if MOBILEGL_BUILD_DISAGGREGATED`, the DEFAULT behaviour is byte-identical to what the
three funnels did on their own, and MG_Remote::InstallPipeSessionFailHook() points it at
SessionFail. The install is in InitServerRoleCommon (MG_Backend/Init.cpp), step 0, because
that is the one function BOTH server roles run - the inproc role and the spawn/TCP session
child - so there is no per-transport install to get wrong.

MGPipeFatalFamily is a two-value boundary enum, NOT MGFatalFamily cast to an integer: the
.def's enumerators are positional and a row inserted in its middle would silently re-point
every hook call at a different family. FatalFunnel.cpp's adapter is the one place the two
vocabularies meet, as a switch with a named default.

The family word stays UnmigratedVerb rather than becoming a new MagmaWireRefusal. Three
reasons: the message strings are required to stay byte-identical (5.2), and the retrace
refusal censuses recorded since P5b count these deaths by that word - a rename invalidates
every `@P7` measurement taken against them; `Magma:` in the detail already separates the
Magma arm from the GLES UnmigratedVerb deaths, so a new word adds no discrimination a grep
does not have; and its wire projection would be ProtocolCorruption, exactly UnmigratedVerb's.
Recorded for the integrator.

Ph slice (2) in the same commit: Server/StagedShadow.h's RequireCovered and
Server/StagedTextureStore.h's RequireLevelBytes were MGLOG_F + bare abort. Both are inside
MG_Remote already, so they call SessionFail directly, message verbatim;
check_include_closure.py's four probes do not reach either header and the gate is green.

Census: 83 abort sites -> 79 over 20 files (five sites retired, one gained for the seam's
no-hook default, which is argued into FUNNEL_SITES), baseline rewritten in this commit - the
downward ratchet's notice is what asked for it. Two gate bugs the widening exposed and this
commit fixes: the SessionFail call regex also matched `MGPipeSessionFail(` (a lookbehind
now), and a funnel's anchor needed a wider window than a marker's (kFunnelAnchorWindow).

Red-once, three ways. Pointing MagmaWireFatal back at its own abort reds the census naming
WireFramebuffer.inc 0 -> 1. Making the seam skip the hook reds both new death cases -
`TheMagmaWireVerbDeathReachesSessionFailAndBumpsTheFaultCount` on the exit code, because it
reads SessionFaultCount() out of a SIGABRT handler and gets 0 instead of 1, and
`TheSeamHandsTheInstalledHookItsFamilyAndItsLineVerbatim` on the line the hook echoed.
This commit is contained in:
2026-09-22 14:01:06 -04:00
parent f141918aa1
commit e1ec69341a
14 changed files with 333 additions and 21 deletions
+95
View File
@@ -8,12 +8,17 @@
// P6 `dl` (CONTRACT-P6 5.2). The Fatal-family table and its projection onto the wire FatalCode.
#include <MG_Pipe/PipeSessionFail.h>
#include <MG_Remote/FatalFamily.h>
#include <MG_Remote/FatalFunnel.h>
#include <MG_Remote/Protocol/generated/protocol_generated.h>
#include <flatbuffers/flatbuffers.h>
#include <gtest/gtest.h>
#include <csignal>
#include <cstdio>
#include <cstdlib>
#include <set>
#include <string>
@@ -29,6 +34,27 @@ namespace {
#undef X
};
#if !defined(_WIN32)
// P7 wave 0. SessionFaultCount() is the only thing the funnel leaves behind that a test can
// read, and SessionFail aborts before returning - so the counter has to be read from INSIDE
// the dying process. A SIGABRT handler turns it into the child's exit code: kFaultExitBase +
// n. "The funnel was reached exactly once" is then a value the parent asserts rather than a
// log line it has to parse, and a seam that bypassed the funnel would exit 0x60, not 0x61.
constexpr int kFaultExitBase = 0x60;
void ReportFaultCountAndExit(int) {
std::_Exit(kFaultExitBase + static_cast<int>(MobileGL::MG_Remote::SessionFaultCount()));
}
// A hook of the TEST's own, to prove what the seam hands across independently of what
// SessionFail then does with it. It echoes and returns; MGPipeSessionFail's contract is that
// it aborts anyway, which is also under test here.
void EchoingHook(MobileGL::MG_Pipe::MGPipeFatalFamily family, const char* line) {
std::fprintf(stderr, "seam-saw family=%u line=%s\n", static_cast<unsigned>(family), line);
std::fflush(stderr);
}
#endif
} // namespace
TEST(FatalFamily, TheTableAndTheEnumAgreeOnCount) {
@@ -84,6 +110,75 @@ TEST(FatalFamily, TheAnchorFamiliesProjectWhereTheirMeaningSays) {
EXPECT_EQ(FatalCodeForFamily(MGFatalFamily::UnmigratedVerb), FatalCode::ProtocolCorruption);
}
// ---------------------------------------------------------------------------------------------
// P7 wave 0: the seam MG_Backend's three Magma wire funnels die through (MG_Pipe/
// PipeSessionFail.h). Until this package they logged and raised their own std::abort(), so
// fifteen `@P7` refusals plus WireBufferLegacyFatal published no SessionFault to the peer, bumped
// no SessionFaultCount() and were invisible to the census gate. These three cases are the
// red-once for that: delete InstallPipeSessionFailHook()'s call in InitServerRoleCommon and the
// first two go red on the exit code, point a funnel back at MGLOG_F + abort and all three do.
TEST(FatalFunnelSeam, InstallingTheHookPointsTheSeamSomewhereOtherThanItsDefault) {
using namespace MobileGL;
ASSERT_EQ(MG_Pipe::MGPipeSessionFailHookInstalled(), nullptr)
<< "the seam was already installed before this case ran; the binary's default is the "
"no-hook one and only a server-role init should change it";
MG_Remote::InstallPipeSessionFailHook();
EXPECT_NE(MG_Pipe::MGPipeSessionFailHookInstalled(), nullptr);
// Restored, so the two cases below start from the same state whichever order they run in.
MG_Pipe::MGPipeInstallSessionFailHook(nullptr);
EXPECT_EQ(MG_Pipe::MGPipeSessionFailHookInstalled(), nullptr);
}
TEST(FatalFunnelSeam, TheMagmaWireVerbDeathReachesSessionFailAndBumpsTheFaultCount) {
#if defined(_WIN32)
GTEST_SKIP() << "the seam's proof forks and reads a signal, which is POSIX only - the same "
"reason MG_Test/Wire's spawn cases are POSIX only";
#else
using namespace MobileGL;
// WireFramebuffer.inc's MagmaWireFatal, character for character - the funnel eleven of the
// fifteen `@P7` refusals die through.
EXPECT_EXIT(
{
std::signal(SIGABRT, &ReportFaultCountAndExit);
MG_Remote::InstallPipeSessionFailHook();
MG_Pipe::MGPipeSessionFail(MG_Pipe::MGPipeFatalFamily::UnmigratedVerb,
"MGPipe: Fatal{UnmigratedVerb, \"Magma:%s\"}",
"unit-probe");
},
::testing::ExitedWithCode(kFaultExitBase + 1),
"Fatal\\{UnmigratedVerb, \"Magma:unit-probe\"\\}")
<< "the Magma wire funnel did not end through SessionFail: either the line is not the "
"site's own any more, or SessionFaultCount() did not move - which is what exit gate "
"S8 reads and what every telemetry consumer of the funnel depends on";
#endif
}
TEST(FatalFunnelSeam, TheSeamHandsTheInstalledHookItsFamilyAndItsLineVerbatim) {
#if defined(_WIN32)
GTEST_SKIP() << "POSIX only, as above";
#else
using namespace MobileGL;
// WireDraw.inc's WireBufferLegacyFatal, character for character. Through a hook of the test's
// own rather than MG_Remote's, so what is under test is the SEAM - that the family enum and
// the fully formatted line cross intact - and not what SessionFail then does with them. The
// hook returns; MGPipeSessionFail must still abort, which is why this is an exit assertion.
EXPECT_EXIT(
{
MG_Pipe::MGPipeInstallSessionFailHook(&EchoingHook);
MG_Pipe::MGPipeSessionFail(
MG_Pipe::MGPipeFatalFamily::RoleViolation,
"MGPipe: Fatal{RoleViolation, \"buffer-legacy-arm\"} (Magma P7 buffer consumer)");
},
::testing::KilledBySignal(SIGABRT),
"seam-saw family=1 line=MGPipe: Fatal\\{RoleViolation, \"buffer-legacy-arm\"\\} "
"\\(Magma P7 buffer consumer\\)")
<< "the seam changed what it hands across: either the family ordinal moved (which would "
"re-point FatalFunnel.cpp's adapter arm at the wrong .def row) or the line is no "
"longer the site's own string, which CONTRACT-P6 5.2 requires to stay byte-identical";
#endif
}
TEST(FatalFamily, TheSessionFaultFrameCarriesTheFamilyAndItsCode) {
// `dl` step three (5.2): the Fatal frame a dying session publishes carries the full family
// WORD beside the coarse code, so a peer names which family ended the session instead of