test_framework: detect failure of bitcoind startup

Replace the `bitcoin-cli -rpcwait` after spawning bitcoind
with our own loop that detects when bitcoind exits prematurely.

And if one node fails to start, stop the others.

This prevents a hang in such a case (see #7463).

1 files changed, 40 insertions, 20 deletions

diff --git a/qa/rpc-tests/test_framework/util.py b/qa/rpc-tests/test_framework/util.py
index 8d4bd52b94..f069c32a60 100644
--- a/qa/rpc-tests/test_framework/util.py
+++ b/qa/rpc-tests/test_framework/util.py
@@ -16,6 +16,7 @@ import shutil
 import subprocess
 import time
 import re
+import errno
 
 from . import coverage
 from .authproxy import AuthServiceProxy, JSONRPCException
@@ -130,11 +131,33 @@ def initialize_datadir(dirname, n):
         f.write("listenonion=0\n")
     return datadir
 
+def rpc_url(i, rpchost=None):
+    return "http://rt:rt@%s:%d" % (rpchost or '127.0.0.1', rpc_port(i))
+
+def wait_for_bitcoind_start(process, url, i):
+    '''
+    Wait for bitcoind to start. This means that RPC is accessible and fully initialized.
+    Raise an exception if bitcoind exits during initialization.
+    '''
+    while True:
+        if process.poll() is not None:
+            raise Exception('bitcoind exited with status %i during initialization' % process.returncode)
+        try:
+            rpc = get_rpc_proxy(url, i)
+            blocks = rpc.getblockcount()
+            break # break out of loop on success
+        except IOError as e:
+            if e.errno != errno.ECONNREFUSED: # Port not yet open?
+                raise # unknown IO error
+        except JSONRPCException as e: # Initialization phase
+            if e.error['code'] != -28: # RPC in warmup?
+                raise # unkown JSON RPC exception
+        time.sleep(0.25)
+
 def initialize_chain(test_dir):
     """
     Create (or copy from cache) a 200-block-long chain and
     4 wallets.
-    bitcoind and bitcoin-cli must be in search path.
     """
 
     if (not os.path.isdir(os.path.join("cache","node0"))
@@ -147,7 +170,6 @@ def initialize_chain(test_dir):
             if os.path.isdir(os.path.join("cache","node"+str(i))):
                 shutil.rmtree(os.path.join("cache","node"+str(i)))
 
-        devnull = open(os.devnull, "w")
         # Create cache directories, run bitcoinds:
         for i in range(4):
             datadir=initialize_datadir("cache", i)
@@ -156,19 +178,15 @@ def initialize_chain(test_dir):
                 args.append("-connect=127.0.0.1:"+str(p2p_port(0)))
             bitcoind_processes[i] = subprocess.Popen(args)
             if os.getenv("PYTHON_DEBUG", ""):
-                print "initialize_chain: bitcoind started, calling bitcoin-cli -rpcwait getblockcount"
-            subprocess.check_call([ os.getenv("BITCOINCLI", "bitcoin-cli"), "-datadir="+datadir,
-                                    "-rpcwait", "getblockcount"], stdout=devnull)
+                print "initialize_chain: bitcoind started, waiting for RPC to come up"
+            wait_for_bitcoind_start(bitcoind_processes[i], rpc_url(i), i)
             if os.getenv("PYTHON_DEBUG", ""):
-                print "initialize_chain: bitcoin-cli -rpcwait getblockcount completed"
-        devnull.close()
+                print "initialize_chain: RPC succesfully started"
 
         rpcs = []
-
         for i in range(4):
             try:
-                url = "http://rt:rt@127.0.0.1:%d" % (rpc_port(i),)
-                rpcs.append(get_rpc_proxy(url, i))
+                rpcs.append(get_rpc_proxy(rpc_url(i), i))
             except:
                 sys.stderr.write("Error connecting to "+url+"\n")
                 sys.exit(1)
@@ -243,17 +261,12 @@ def start_node(i, dirname, extra_args=None, rpchost=None, timewait=None, binary=
     args = [ binary, "-datadir="+datadir, "-server", "-keypool=1", "-discover=0", "-rest", "-mocktime="+str(get_mocktime()) ]
     if extra_args is not None: args.extend(extra_args)
     bitcoind_processes[i] = subprocess.Popen(args)
-    devnull = open(os.devnull, "w")
     if os.getenv("PYTHON_DEBUG", ""):
-        print "start_node: bitcoind started, calling bitcoin-cli -rpcwait getblockcount"
-    subprocess.check_call([ os.getenv("BITCOINCLI", "bitcoin-cli"), "-datadir="+datadir] +
-                          _rpchost_to_args(rpchost)  +
-                          ["-rpcwait", "getblockcount"], stdout=devnull)
+        print "start_node: bitcoind started, waiting for RPC to come up"
+    url = rpc_url(i, rpchost)
+    wait_for_bitcoind_start(bitcoind_processes[i], url, i)
     if os.getenv("PYTHON_DEBUG", ""):
-        print "start_node: calling bitcoin-cli -rpcwait getblockcount returned"
-    devnull.close()
-    url = "http://rt:rt@%s:%d" % (rpchost or '127.0.0.1', rpc_port(i))
-
+        print "start_node: RPC succesfully started"
     proxy = get_rpc_proxy(url, i, timeout=timewait)
 
     if COVERAGE_DIR:
@@ -267,7 +280,14 @@ def start_nodes(num_nodes, dirname, extra_args=None, rpchost=None, binary=None):
     """
     if extra_args is None: extra_args = [ None for i in range(num_nodes) ]
     if binary is None: binary = [ None for i in range(num_nodes) ]
-    return [ start_node(i, dirname, extra_args[i], rpchost, binary=binary[i]) for i in range(num_nodes) ]
+    rpcs = []
+    try:
+        for i in range(num_nodes):
+            rpcs.append(start_node(i, dirname, extra_args[i], rpchost, binary=binary[i]))
+    except: # If one node failed to start, stop the others
+        stop_nodes(rpcs)
+        raise
+    return rpcs
 
 def log_filename(dirname, n_node, logname):
     return os.path.join(dirname, "node"+str(n_node), "regtest", logname)