8319859: Better symbol storage

Reviewed-by: rhalade, coleenp, matsaave, iklam

src/hotspot/share/classfile/symbolTable.cpp

@@ -344,8 +344,19 @@ Symbol* SymbolTable::lookup_common(const char* name,
   return sym;
 }
 
+// Symbols should represent entities from the constant pool that are
+// limited to 64K in length, but usage errors creep in allowing Symbols
+// to be used for arbitrary strings. For debug builds we will assert if
+// a string is too long, whereas product builds will truncate it.
 Symbol* SymbolTable::new_symbol(const char* name, int len) {
-  assert(len <= Symbol::max_length(), "sanity");
+  assert(len <= Symbol::max_length(),
+         "String length %d exceeds the maximum Symbol length of %d", len, Symbol::max_length());
+  if (len > Symbol::max_length()) {
+    warning("A string \"%.80s ... %.80s\" exceeds the maximum Symbol "
+            "length of %d and has been truncated", name, (name + len - 80), Symbol::max_length());
+    len = Symbol::max_length();
+  }
+
   unsigned int hash = hash_symbol(name, len, _alt_hash);
   Symbol* sym = lookup_common(name, len, hash);
   if (sym == nullptr) {
@@ -485,6 +496,7 @@ void SymbolTable::new_symbols(ClassLoaderData* loader_data, const constantPoolHa
   for (int i = 0; i < names_count; i++) {
     const char *name = names[i];
     int len = lengths[i];
+    assert(len <= Symbol::max_length(), "must be - these come from the constant pool");
     unsigned int hash = hashValues[i];
     assert(lookup_shared(name, len, hash) == nullptr, "must have checked already");
     Symbol* sym = do_add_if_needed(name, len, hash, is_permanent);
@@ -494,6 +506,7 @@ void SymbolTable::new_symbols(ClassLoaderData* loader_data, const constantPoolHa
 }
 
 Symbol* SymbolTable::do_add_if_needed(const char* name, int len, uintx hash, bool is_permanent) {
+  assert(len <= Symbol::max_length(), "caller should have ensured this");
   SymbolTableLookup lookup(name, len, hash);
   SymbolTableGet stg;
   bool clean_hint = false;

src/hotspot/share/oops/symbol.cpp

@@ -54,6 +54,7 @@ uint32_t Symbol::pack_hash_and_refcount(short hash, int refcount) {
 }
 
 Symbol::Symbol(const u1* name, int length, int refcount) {
+  assert(length <= max_length(), "SymbolTable should have caught this!");
   _hash_and_refcount =  pack_hash_and_refcount((short)os::random(), refcount);
   _length = (u2)length;
   // _body[0..1] are allocated in the header just by coincidence in the current

src/hotspot/share/oops/symbol.hpp

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1997, 2023, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1997, 2024, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -130,6 +130,7 @@ class Symbol : public MetaspaceObj {
     return (int)heap_word_size(byte_size(length));
   }
 
+  // Constructor is private for use only by SymbolTable.
   Symbol(const u1* name, int length, int refcount);
 
   static short extract_hash(uint32_t value)   { return (short)(value >> 16); }