requireFill: exempt the first run that happened, not the first clock tick.

The gate skipped the analysis block when the observed container held too
little, except on the first run - and "first run" was read off lastAnalysis,
which holds the experiment time. That is exactly zero as long as an
experiment has never been started, so before the first start every pass
counted as the first one and the gate never engaged; the golden vector for
the ruled semantics (spec/analysis.yml requireFill, decided 2026-08-24) is
pinned in precisely that state and saw the block run twice.

A pass that actually ran now sets analysisRan, and stopAllIO clears it again
so the first run after the next start stays exempt. While measuring nothing
changes: the experiment time was non-zero there, so the old test agreed with
the new one. Before the first start, analysis on user input is now gated
from the second pass on, which is what the rule says.

The vector runs unskipped again: 117 of 117.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Author: Staacks

app/src/main/java/de/rwth_aachen/phyphox/PhyphoxExperiment.java

@@ -124,6 +124,7 @@ public List<Link> getHighlightedLinks() {
     double analysisSleep = 0.; //Pause between analysis cycles. At 0 analysis is done as fast as possible.
     DataBuffer analysisDynamicSleep = null;
     double lastAnalysis = 0.0; //This variable holds the system time of the moment the last analysis process finished. This is necessary for experiments, which do analysis after given intervals
+    boolean analysisRan = false; //Whether an analysis pass has run since the experiment was opened or (re)started. The first run is exempt from the requireFill gate (spec/analysis.yml, decided 2026-08-24), and this cannot be read off lastAnalysis: that holds the experiment time, which is exactly zero as long as the experiment has never been started.
     double analysisTime; //This variable holds the experiment time of the moment the current analysis process started.
     double analysisLinearTime; //Same with the current system time
     boolean analysisOnUserInput = false; //Do the data analysis only if there is fresh input from the user.
@@ -327,7 +328,7 @@ public void processAnalysis(boolean measuring) {
         } else
             cycle = 0;
 
-        if (requireFill != null && lastAnalysis != 0) {
+        if (requireFill != null && analysisRan) {
             int threshold = requireFillThreshold;
             if (requireFillDynamic != null && requireFillDynamic.getFilledSize() > 0)
                 threshold = (int)requireFillDynamic.value;
@@ -392,6 +393,7 @@ public void processAnalysis(boolean measuring) {
         recordingUsed = true;
         newData = true; //We have fresh data to present.
         lastAnalysis = experimentTimeReference.getExperimentTime(); //Remember when we were done this time
+        analysisRan = true; //A run that was gated above does not get here, so this marks a pass that actually happened
     }
 
     //called by the main loop after everything is processed. Here we have to send all the analysis results to the appropriate views
@@ -441,6 +443,7 @@ public void stopAllIO() {
             experimentTimeReference.registerEvent(ExperimentTimeReference.TimeMappingEvent.PAUSE);
         event = experimentTimeReference.getLastMapping();
         lastAnalysis = 0.0;
+        analysisRan = false; //The first run after the next start is exempt from the requireFill gate again
 
         //Recording
         if (audioRecord != null && audioRecord.getState() == AudioRecord.STATE_INITIALIZED)

app/src/test/java/de/rwth_aachen/phyphox/AnalysisGoldenVectorTest.java

@@ -41,24 +41,6 @@ public class AnalysisGoldenVectorTest {
 
     private static final String VECTORS = "analysis/vectors";
 
-    //Cases whose failure has been reported and is waiting for a decision, skipped with the
-    //finding so the rest of the corpus stays useful. An entry here is a report, not a fix: it
-    //states what the app does and why that differs from what the case pins, and it goes away
-    //when the divergence is resolved. Never add one to make a red case quiet.
-    private static final Map<String, String> REPORTED = new LinkedHashMap<String, String>() {{
-        put("execution/requirefill-first-run-exempt.phyphox",
-                "the app exempts the first run through the experiment clock - processAnalysis "
-                        + "gates on \"requireFill != null && lastAnalysis != 0\", and lastAnalysis "
-                        + "holds the experiment time, which stays exactly 0 while an experiment "
-                        + "has never been started. So in the never-started state this case pins, "
-                        + "EVERY pass is exempt and the second cycle appends again (out holds "
-                        + "1, 2, 1, 2). Once the experiment has been started the gate behaves as "
-                        + "ruled. Fixing it means keying the exemption on a run having happened "
-                        + "rather than on the clock, which changes what analysis-on-user-input "
-                        + "does before the first start - a production change, so reported rather "
-                        + "than made here.");
-    }};
-
     private final String relativePath;
 
     public AnalysisGoldenVectorTest(String relativePath) {
@@ -97,10 +79,6 @@ public void matchesGoldenVector() throws Exception {
                             + " > supported " + PhyphoxFile.phyphoxFileVersion + " - skipped.",
                     CorpusTestEnvironment.versionAtMostSupported(declared));
 
-        for (Map.Entry<String, String> reported : REPORTED.entrySet())
-            assumeTrue("Reported to phyphox-docs: " + reported.getValue(),
-                    !relativePath.endsWith(reported.getKey()));
-
         JSONObject expected = readJson(new File(corpus,
                 relativePath.substring(0, relativePath.length() - ".phyphox".length()) + ".expected.json"));