Persistence : fix hang when SP watermark cannot be located

The methodology for loading persistence relies on walking backwards through persistence files until either a) we run out of persistence files or b) we run out of blocks.

This is done via a while loop, but the while is based on persistedBlocks.size() which is only changed if persistedBlocks.find() is successful in locating the block.  Thus if persistedBlocks.find() is unsuccesful (for example in the case of failed persistence loads) then the while loop will keep iterating - subtracting a block from curTip and trying to load state from persistence files again - over and over all the way back to the genesis block.

This manifests in an apparently hung client.

This commit adds an exit from the while after MAX_STATE_HISTORY+1 load attempts to prevent the issue.

src/mastercore.cpp

@@ -2042,8 +2042,11 @@ static int load_most_relevant_state()
   // using the SP's watermark after its fixed-up as the tip
   // walk backwards until we find a valid and full set of persisted state files
   // for each block we discard, roll back the SP database
+  // Note: to avoid rolling back all the way to the genesis block (which appears as if client is hung) abort after MAX_STATE_HISTORY attempts
   CBlockIndex const *curTip = spBlockIndex;
-  while (NULL != curTip && persistedBlocks.size() > 0) {
+  int abortRollBackBlock;
+  if (curTip != NULL) abortRollBackBlock = curTip->nHeight - MAX_STATE_HISTORY+1;
+  while (NULL != curTip && persistedBlocks.size() > 0 && curTip->nHeight > abortRollBackBlock) {
     if (persistedBlocks.find(spBlockIndex->GetBlockHash()) != persistedBlocks.end()) {
       int success = -1;
       for (int i = 0; i < NUM_FILETYPES; ++i) {
@@ -2421,6 +2424,8 @@ int mastercore_init()
   if (readPersistence())
   {
     nWaterlineBlock = load_most_relevant_state();
+    PrintToConsole("Loading persistent state: %s\n", nWaterlineBlock>0 ?
+        "OK" : "FAILED (this is normal if this is the first time OmniCore is being run)");
     if (nWaterlineBlock < 0) {
       // persistence says we reparse!, nuke some stuff in case the partial loads left stale bits
       clear_all_state();