[lldb] Store *signed* ranges in lldb_private::Block (llvm#120224)

This is to support functions whose entry points aren't their lowest
address

(https://discourse.llvm.org/t/rfcish-support-for-discontinuous-functions/83244).
The alternative is to keep blocks relative to the lowest address, but
then introduce a separate concept for the function entry point, which I
think would be more confusing.

This patch just changes the type signedness, it doesn't create any
negative offsets yet. Since combining values with different signs can
sometimes produce unexpected results, and since this is the first use of
RangeVector with a signed type, I'm adding a test to verify that at
least the core functionality works correctly.

lldb/include/lldb/Symbol/Block.h

@@ -40,7 +40,7 @@ namespace lldb_private {
 /// blocks.
 class Block : public UserID, public SymbolContextScope {
 public:
-  typedef RangeVector<uint32_t, uint32_t, 1> RangeList;
+  typedef RangeVector<int32_t, uint32_t, 1> RangeList;
   typedef RangeList::Entry Range;
 
   // Creates a block representing the whole function. Only meant to be used from

lldb/unittests/Utility/RangeMapTest.cpp

@@ -12,6 +12,29 @@
 
 using namespace lldb_private;
 
+TEST(RangeVector, SignedBaseType) {
+  using RangeVector = RangeVector<int32_t, uint32_t>;
+  using Entry = RangeVector::Entry;
+
+  RangeVector V;
+  V.Append(10, 5);
+  V.Append(-3, 6);
+  V.Append(-10, 3);
+  V.Sort();
+  EXPECT_THAT(V,
+              testing::ElementsAre(Entry(-10, 3), Entry(-3, 6), Entry(10, 5)));
+  Entry e = *V.begin();
+  EXPECT_EQ(e.GetRangeBase(), -10);
+  EXPECT_EQ(e.GetByteSize(), 3u);
+  EXPECT_EQ(e.GetRangeEnd(), -7);
+  EXPECT_TRUE(e.Contains(-10));
+  EXPECT_TRUE(e.Contains(-8));
+  EXPECT_FALSE(e.Contains(-7));
+  EXPECT_TRUE(e.Union(Entry(-8, 2)));
+  EXPECT_EQ(e, Entry(-10, 4));
+  EXPECT_EQ(e.Intersect(Entry(-7, 3)), Entry(-7, 1));
+}
+
 TEST(RangeVector, CombineConsecutiveRanges) {
   using RangeVector = RangeVector<uint32_t, uint32_t>;
   using Entry = RangeVector::Entry;