From 485e89f7a5f06cb0e85ac38705e35f3b3f966476 Mon Sep 17 00:00:00 2001
From: Seth Hillbrand <seth@kipro-pcb.com>
Date: Fri, 4 Mar 2022 10:45:49 -0800
Subject: [PATCH] Finally found the intermittant QA crash

When loading footprints/zones in parallel, we cannot reference a static
vector that gets cleared by other threads.

It is unclear why this element was ever static to begin with (premature
optimization?) but it has been this way for a long time.  We never
noticed until we threaded load cycles, which gave the hatch a chance to
override itself and cause crashes

Fixes https://gitlab.com/kicad/code/kicad/issues/9888

(cherry picked from commit 8218f9ab44e9aa1ce8ec6ce4ce9b7d6b34eadc7c)
---
 pcbnew/zone.cpp | 41 ++++++-----------------------------------
 1 file changed, 6 insertions(+), 35 deletions(-)

diff --git a/pcbnew/zone.cpp b/pcbnew/zone.cpp
index 6290b36c4d..f897006309 100644
--- a/pcbnew/zone.cpp
+++ b/pcbnew/zone.cpp
@@ -971,54 +971,25 @@ void ZONE::HatchBorder()
     min_a += offset;
 
     // loop through hatch lines
-    #define MAXPTS 200      // Usually we store only few values per one hatch line
-                            // depending on the complexity of the zone outline
-
-    static std::vector<VECTOR2I> pointbuffer;
-    pointbuffer.clear();
-    pointbuffer.reserve( MAXPTS + 2 );
+    std::vector<VECTOR2I> pointbuffer;
+    pointbuffer.reserve( 256 );
 
     for( int a = min_a; a < max_a; a += spacing )
     {
-        // get intersection points for this hatch line
-
-        // Note: because we should have an even number of intersections with the
-        // current hatch line and the zone outline (a closed polygon,
-        // or a set of closed polygons), if an odd count is found
-        // we skip this line (should not occur)
         pointbuffer.clear();
 
         // Iterate through all vertices
         for( auto iterator = m_Poly->IterateSegmentsWithHoles(); iterator; iterator++ )
         {
             double x, y;
-            bool   ok;
 
             SEG segment = *iterator;
 
-            ok = FindLineSegmentIntersection( a, slope, segment.A.x, segment.A.y, segment.B.x,
-                                              segment.B.y, x, y );
-
-            if( ok )
-            {
-                VECTOR2I point( KiROUND( x ), KiROUND( y ) );
-                pointbuffer.push_back( point );
-            }
-
-            if( pointbuffer.size() >= MAXPTS ) // overflow
-            {
-                wxASSERT( 0 );
-                break;
-            }
+            if( FindLineSegmentIntersection( a, slope, segment.A.x, segment.A.y, segment.B.x,
+                    segment.B.y, x, y ) )
+                pointbuffer.emplace_back( KiROUND( x ), KiROUND( y ) );
         }
 
-        // ensure we have found an even intersection points count
-        // because intersections are the ends of segments
-        // inside the polygon(s) and a segment has 2 ends.
-        // if not, this is a strange case (a bug ?) so skip this hatch
-        if( pointbuffer.size() % 2 != 0 )
-            continue;
-
         // sort points in order of descending x (if more than 2) to
         // ensure the starting point and the ending point of the same segment
         // are stored one just after the other.
@@ -1026,7 +997,7 @@ void ZONE::HatchBorder()
             sort( pointbuffer.begin(), pointbuffer.end(), sortEndsByDescendingX );
 
         // creates lines or short segments inside the complex polygon
-        for( unsigned ip = 0; ip < pointbuffer.size(); ip += 2 )
+        for( size_t ip = 0; ip + 1 < pointbuffer.size(); ip += 2 )
         {
             int dx = pointbuffer[ip + 1].x - pointbuffer[ip].x;