[Fix] (MG_Remote, GLImpl): gate the transform-feedback reconciliation on the transport like every other new site, and give the unwritten writeback wait a named Fatal instead of an empty body

MobileGL/MG_Impl/GLImpl/Drawing/GL_Drawing.cpp

@@ -1324,7 +1324,13 @@ namespace MobileGL::MG_Impl::GLImpl {
             // fence introduces. Under split it pays the reconciliation itself, which is the
             // same cost the fence used to charge every caller - here charged only to the
             // capture shapes that actually need reordering.
-            buffer->SyncGpuWrites();
+            //
+            // TRANSPORT-GATED LIKE EVERY OTHER NEW SITE (D-J). Without the test this fires in
+            // a build-split lane running MOBILEGL_TRANSPORT=monolith on Magma - whose
+            // BeginXfbCaptureForDraw does mark the capture targets - where the fence at the
+            // caller still runs, so the readback it emits is pure new work on the monolith
+            // path and integration-gpu cannot see it.
+            if (MG_Config::Transport != MG_Config::TransportMode::Monolith) buffer->SyncGpuWrites();
 #endif
             const Range1D range = bindingPoint.GetRange();
             const Uint8* mapped = buffer->MappedData();

MobileGL/MG_Remote/Client/GpuWritePending.cpp

@@ -15,6 +15,9 @@
 #include <MG_State/GLState/Core.h>
 #include <MG_State/GLState/TextureState/TextureObjectBuffer.h>
 #include <MG_State/GLState/TextureState/TextureState.h>
+#include <MG_Util/Debug/Log.h>
+
+#include <cstdlib>
 
 namespace MobileGL::MG_Remote::Client {
 
@@ -167,16 +170,31 @@ namespace MobileGL::MG_Remote::Client {
     }
 
     void AwaitBufferWriteback(BufferObject& buffer) {
-        (void)buffer;
         // THE WAIT IS THE BARRIER'S WAIT (R-3). The reply-slot id IS the record's seq, so
         // "appliedSeq reached my readback" and "my answer is back" are one condition, and
-        // ClientSession::EmitAndWait has already paid for it by the time the emitter returns.
+        // ClientSession::EmitAndWait is what pays for it. With no session - a build-split lane
-        // With no session - a build-split lane running monolith, and every unit case - the
+        // running monolith, and every unit case - the emission WAS the application,
-        // emission WAS the application, synchronously, so the writeback has already landed
+        // synchronously, so the writeback has already landed and there is nothing to wait for.
-        // and there is nothing to wait for. Spelling that as "return" rather than as a loop
+        // Spelling that as "return" rather than as a loop is deliberate: a loop here would be
-        // is deliberate: a loop here would be a hang in exactly that configuration, which is
+        // a hang in exactly that configuration, which is the configuration every gate lane
-        // the configuration every gate lane runs.
+        // runs.
         if (ClientSession::Active() == nullptr) return;
+
+        // AND THE OTHER ARM IS A NAMED FATAL, NOT AN EMPTY BODY. A session exists, so the
+        // apply side is no longer synchronous, and if the flag is still set the shadow this
+        // caller is about to read is STALE - which is the whole failure the third state was
+        // introduced to stop. An empty body here would make that failure silent and would let
+        // s1/c1 land a session without noticing that nobody ever wrote the wait; a stub that
+        // aborts by name is the house shape for exactly this (EmitTables.cpp's
+        // UnmigratedVerbFatal), and it is what gives the hole a red spelling before the
+        // transport arrives.
+        if (!buffer.HasOutstandingGpuWrite()) return;
+        MGLOG_F("MGPipe: Fatal{UnimplementedWritebackWait} - a ClientSession is active and buffer %u "
+                "still has an outstanding GPU write after its readback was emitted. The wait is "
+                "ClientSession::EmitAndWait's (R-3: the reply slot id IS the record seq); P5 package "
+                "b1 landed the third state and s1/c1 own the wait itself.",
+                buffer.GetExternalIndex());
+        std::abort();
     }
 
 } // namespace MobileGL::MG_Remote::Client

MobileGL/MG_Remote/Client/GpuWritePending.h

@@ -41,6 +41,13 @@
 //     ~0ull) and marks the capture targets, for the same reason: the wait exists only so that
 //     a later MapBuffer sees real results, which is precisely what the flag is for.
 //
+// WHAT NO ROW COVERS, WRITTEN DOWN SO THE NEXT READER DOES NOT HAVE TO ASK. The set is built
+// from the APPLICATION's bindings, so it says nothing about a backend's own scratch buffers -
+// Espryt's converted-vertex-stream and primitive-restart substitution buffers, Magma's UBO
+// ring. None of those has a MarkGpuWritten today either, so the client set is no NARROWER than
+// monolith's and this is not a regression; it is a standing hole in both, and it stays one
+// until the phase that migrates the backend's own allocations.
+//
 // GATED ON THE TRANSPORT, like everything else in this package: on the monolith path the six
 // backend sites still run and a second marker would be new behaviour (D-J), and rows 4 and 5
 // would remove a stall monolith is entitled to keep.

MobileGL/MG_Remote/Client/PersistentMapTracker.cpp

@@ -123,9 +123,22 @@ namespace MobileGL::MG_Remote::Client {
         // parses them today is that the negative control needs a spelling before the thing
         // it controls exists. Falling back would make `MOBILEGL_IPC_ADOPT_TIER=0` look like
         // a working T0 run and silently produce pmap bytes it must not produce.
-        MGLOG_F("MGPipe: MOBILEGL_IPC_ADOPT_TIER=%u names an adoption tier P11 implements and P5 "
+        // 0 and 1 are the two CONTRACT §5 promises - a real cross-process shared mapping and a
-                "does not; P5 runs at T2 (emulate) only.",
+        // server-side staging map - and they name P11. Anything else is not a tier at all, and
-                static_cast<unsigned>(tier));
+        // saying "P11 implements it" of a 7 would be a lie the operator then repeats. Both die
+        // here rather than at parse, which is late: the abort lands at the first
+        // map_persistent, so a mis-set run gets through EGL bring-up and a frame of setup
+        // first. Moving it to the parse means a knob-validity rule in ConfigLoader, which is
+        // c0's file; filed for the integrator rather than taken here.
+        if (tier <= 1) {
+            MGLOG_F("MGPipe: MOBILEGL_IPC_ADOPT_TIER=%u names adoption tier T%u, which P11 implements "
+                    "and P5 does not; P5 runs at T2 (emulate) only.",
+                    static_cast<unsigned>(tier), static_cast<unsigned>(tier));
+        } else {
+            MGLOG_F("MGPipe: MOBILEGL_IPC_ADOPT_TIER=%u is not an adoption tier; the only values are 0 "
+                    "and 1 (P11) and 2 (emulate, the P5 default).",
+                    static_cast<unsigned>(tier));
+        }
         std::abort();
     }