DFX6xxx — Niagara compilation

8 codes in this family (5 error, 2 warning, 1 info). The message is the format string as the compiler holds it: %s, %d and %c are substituted at runtime.

DFX6001

Cause. A Niagara compile error, mapped back to the source line of the module that raised it.

Fix. The message is Niagara's own. An unresolved attribute usually means a module that writes it is missing or runs later in the stack than the module that reads it.

DFX6002

Cause. A Niagara compile warning, mapped back to source.

Fix. Read the message. Warnings from a dependency are marked as such.

DFX6003

Cause. A Niagara stack issue at error level -- an unmet module dependency, most often. Stack issues are only readable where Slate exists, so these appear in the editor and in the corpus suite but not in a headless build (GetStackIssues is not headless-safe).

Fix. Add the module the dependency names. ScaleSpriteSize requiring UpdateAge (provided by ParticleState) is the common one -- without it NormalizedAge never advances and every age-driven curve evaluates at zero.

DFX6004

Cause. A Niagara stack issue at warning level. A deprecated module reports here.

Fix. The message usually names the replacement asset.

DFX6005

Cause. The system's compile did not reach a successful state. Individual errors are reported as DFX6001 above this.

Fix. Fix the errors above. A status of Dirty with no errors means the compile did not finish.

DFX6006

Cause. A .dfm's body is not valid HLSL once lowered. The Niagara message follows.

Fix. Read the translator error. Cannot access field 'X' of structure means a namespaced reference did not resolve -- Particles.* has to be a known attribute or declared with a type in the body (DFX3046).

DFX6007

Cause. Attached to a DFX6001 "read before being set": nothing in the source text writes the parameter, so no link write created it and there is no graph parameter for a default to sit on.

Check the asset before the effect. The first asset to carry this message, NS_Spawn_Ground_Root, was ruled a defective effect on the evidence that its export had no writer --- and the export was the lossy step. A suppressed static switch several stacks away had been dropped by the decompiler, the rebuild compiled the other branch, and the missing write was a casualty with an unrelated name. The asset wrote the parameter all along; once switches export unconditionally (2026-08-11) it builds clean on both engines. That ruling is void, and the shape of the mistake is the house rule it produced: proving "the asset does not contain X" takes asset-level evidence (PkgInfo, a reflection walk, asset-diff), never an export product.

Fix. First re-export the asset with a current build --- if the message goes away, the source was missing something the exporter used to drop, and nothing was ever wrong with the effect. If the source genuinely is the whole story (hand-written, or the asset really has no writer), fix it there: write the parameter before whatever reads it --- for an Emitter.<Module>.<Output> name, move the module that produces it above its readers --- or drop the read. Building on MoonEngine makes the message go away by supplying the type's zero, which is not the same as making the effect work.

Why DreamFX does not just create the parameter. It could: FGraphSurgeon::AddParameter already creates one by reflection for the .dfm path. Deliberately not wired in --- reflection parameter creation in the .dfs path would invert the layer dependency and paper over real source defects on every engine.

Not a reason to distrust the build. It is an explanation attached to an error, never an error of its own.

DFX6008

Cause. The build finished its Niagara compile, and the engine's own synchronization test then said some script's stored VM was not compiled from the graphs as this build left them. That should be impossible with the build forcing its final compile, which is exactly why it is an error: either a write path mutated a graph after the compile was issued, or the force was lost somewhere. The check exists because this class of asset has shipped before --- NE_C of NS_Spawn_Teleport_Root carried bytecode from an all-Unset-switches era under the compile id of the correct graph, and simulated the wrong branches for days while every text-level check agreed with the source (2026-08-11).

Fix. Not in the source --- the message names a DreamFX pipeline defect. Rebuild once with -Force; if it recurs, report it with the source file and the named scripts. The stale scripts list says which emitter and which stage to look at.