Fix bantu-verbal G5 second-sweep grammar bugs

The session that added the mprRRealTest MorphologicalPhonologicalRuleFeature
appended it after the FeatureGroup elements, violating the DTD's element
ordering and crashing the loader. Reordered it before the groups.

Also, snG5's StemName Region carried an AssignedHeadFeatures constraint
(featTense=symPres) that nothing in the daleh/daheh derivation ever assigns,
so RootAllomorph's StemName.IsRequiredMatch check failed root selection
before the intended SynthesisAffixProcessRule.RequiredStemName check could
even run. Dropped the head-feature constraint so the region only requires
posV7, restoring the intended failure/success split (daleh parses, daheh
does not).

Self-check 15/15, coverage 24/24 in-scope constructs, HC tests 67/67,
csharpier clean.

Author: johnml1135

conformance/coverage.csv

@@ -118,6 +118,12 @@ Velamba,dabepez,DAB+NUMPL+AGR|dabepez,Syntactic feature agreement (RequiredHeadF
 Velamba,dabez,DAB+AGR|dabez,Syntactic feature agreement (RequiredHeadFeatures/OutputHeadFeatures/RequiredFootFeatures/OutputFootFeatures)
 Velamba,kes,KES|kes,Syntactic feature agreement (RequiredHeadFeatures/OutputHeadFeatures/RequiredFootFeatures/OutputFootFeatures)
 Velamba,kesez,,Syntactic feature agreement (RequiredHeadFeatures/OutputHeadFeatures/RequiredFootFeatures/OutputFootFeatures)
+Velamba,siliz,SIL+RECURSUF|siliz,Ordinary/realizational rule constraints (MaxApplicationCount/RequiredStemName/Blockable)
+Velamba,siliziz,,Ordinary/realizational rule constraints (MaxApplicationCount/RequiredStemName/Blockable)
+Velamba,daleh,DAL+STEMREQ|daleh,Ordinary/realizational rule constraints (MaxApplicationCount/RequiredStemName/Blockable)
+Velamba,daheh,,Ordinary/realizational rule constraints (MaxApplicationCount/RequiredStemName/Blockable)
+Velamba,zimed,,Ordinary/realizational rule constraints (MaxApplicationCount/RequiredStemName/Blockable)
+Velamba,zamed,,Ordinary/realizational rule constraints (MaxApplicationCount/RequiredStemName/Blockable)
 Vetruna,mun,STEM|mun,Stem names
 Vetruna,muno,,Stem names
 Vetruna,muns,,Stem names

conformance/languages/bantu-verbal/grammar.xml

@@ -112,24 +112,29 @@
       <MorphologicalPhonologicalRuleFeature id="mprConcB">ConcB</MorphologicalPhonologicalRuleFeature>
       <MorphologicalPhonologicalRuleFeature id="mprConcC">ConcC</MorphologicalPhonologicalRuleFeature>
       <MorphologicalPhonologicalRuleFeature id="mprConcD">ConcD</MorphologicalPhonologicalRuleFeature>
-      <MorphologicalPhonologicalRuleFeatureGroup isActive="yes" matchType="any" outputType="overwrite" features="mprExtA mprExtB"><Name>ExtGroup</Name></MorphologicalPhonologicalRuleFeatureGroup>
-      <MorphologicalPhonologicalRuleFeatureGroup isActive="yes" matchType="all" outputType="overwrite" features="mprConcA mprConcB"><Name>AllConc</Name></MorphologicalPhonologicalRuleFeatureGroup>
-      <MorphologicalPhonologicalRuleFeatureGroup isActive="yes" matchType="any" outputType="overwrite" features="mprConcC mprConcD"><Name>AnyConc</Name></MorphologicalPhonologicalRuleFeatureGroup>
       <!-- G5 (code-coverage-driven, second sweep): gates mrRRealTest below (SynthesisRealizationalAffixProcessRule),
            a hard set-membership check so that rule can never spuriously apply to any pre-existing
            word (RequiredHeadFeatures/RealizationalFeatures unify vacuously against a feature no
            other entry ever sets, per the mrAgr/dabez finding above, but requiredMPRFeatures is a
            genuine set-membership gate, not unification). -->
       <MorphologicalPhonologicalRuleFeature id="mprRRealTest">RRealTest</MorphologicalPhonologicalRuleFeature>
+      <MorphologicalPhonologicalRuleFeatureGroup isActive="yes" matchType="any" outputType="overwrite" features="mprExtA mprExtB"><Name>ExtGroup</Name></MorphologicalPhonologicalRuleFeatureGroup>
+      <MorphologicalPhonologicalRuleFeatureGroup isActive="yes" matchType="all" outputType="overwrite" features="mprConcA mprConcB"><Name>AllConc</Name></MorphologicalPhonologicalRuleFeatureGroup>
+      <MorphologicalPhonologicalRuleFeatureGroup isActive="yes" matchType="any" outputType="overwrite" features="mprConcC mprConcD"><Name>AnyConc</Name></MorphologicalPhonologicalRuleFeatureGroup>
     </MorphologicalPhonologicalRuleFeatures>
     <!-- G5 (code-coverage-driven, second sweep): snG5 gates mrStemReq below
          (SynthesisAffixProcessRule.RequiredStemName), previously zero coverage for the FAILURE case
-         (a candidate root without the required stem name). -->
+         (a candidate root without the required stem name). Region carries no AssignedHeadFeatures:
+         RootAllomorph.CheckAllomorphConstraints (StemName.IsRequiredMatch) unifies the region against
+         the word's SyntacticFeatureStruct at root-selection time, and posV7 alone (already on the
+         region via partsOfSpeech) is all we want to require; adding a tense feature there would gate
+         root selection itself on a feature nothing in this derivation ever assigns, masking the
+         RequiredStemName rule-level check (SynthesisAffixProcessRule) this fixture actually targets. -->
     <StemNames>
       <StemName id="snG5" partsOfSpeech="posV7">
         <Name>G5Stem</Name>
         <Regions>
-          <Region><AssignedHeadFeatures><FeatureValue feature="featTense" symbolValues="symPres" /></AssignedHeadFeatures></Region>
+          <Region></Region>
         </Regions>
       </StemName>
     </StemNames>

conformance/rules.csv

@@ -54,10 +54,13 @@ Velamba,mrExtX,xkib;yxkib;xkibz
 Velamba,mrExtY,yxkib
 Velamba,mrFV,andika;andikisha;andikila;andikishila;andikilisha
 Velamba,mrNumPl,dabep;dabepez
+Velamba,mrRecurSuf,siliz;!siliziz
 Velamba,mrRedup,kimbiakimbia
+Velamba,mrStemReq,daleh;!daheh
 Velamba,mrTense,limape
 Velamba,prNasalMutation,mba;!mpa
 Velamba,rrPast,limaile;!limapeile
+Velamba,rrRRealTest,!zimed;!zamed
 Vetruna,mrAblaut,sang
 Vetruna,mrCircumfixGeT,gelobt;gelobth
 Vetruna,mrCompoundConstrained,mitlav;ficlav;!seclav;!genlav;!mitlavlav