bonds: use rung <=> run -G that hurts my pinky
[tests.git] / system / TestMain.py
index b0db61f..19a934d 100755 (executable)
-#!/usr/bin/env python
-# $Id$
+#!/usr/bin/python3 -u
 
+# Thierry Parmentelat <thierry.parmentelat@inria.fr>
+# Copyright (C) 2010 INRIA 
+#
 import sys, os, os.path
-from optparse import OptionParser
+from argparse import ArgumentParser
 import traceback
-from time import strftime
 import readline
+import glob
+from datetime import datetime
 
 import utils
-from TestPlc import TestPlc
+from TestPlc import TestPlc, Ignored
+from TestBonding import TestBonding, onelab_bonding_spec
 from TestSite import TestSite
 from TestNode import TestNode
+from macros import sequences
 
-class TestMain:
+# add $HOME in PYTHONPATH so we can import LocalSubstrate.py
+sys.path.append(os.environ['HOME'])
+import LocalSubstrate
+
+class Step:
+
+    natives = TestPlc.__dict__
 
-    subversion_id = "$Id$"
+    def display (self):
+        return self.name.replace('_', '-')
+    def internal (self):
+        return self.name.replace('-', '_')
+
+    def __init__ (self, name):
+        self.name = name
+        # a native step is implemented as a method on TestPlc
+        self.native = self.internal() in Step.natives
+        if self.native:
+            self.method = Step.natives[self.internal()]
+        else:
+            try:
+                self.substeps = sequences[self.internal()]
+            except Exception as e:
+                print("macro step {} not found in macros.py ({}) - exiting".format(self.display(),e))
+                raise
+
+    def print_doc (self, level=0):
+        tab = 32
+        trail = 8
+        if self.native:
+            start = level*' ' + '* '
+            # 2 is the len of '* '
+            width = tab - level - 2
+            format = "%%-%ds" % width
+            line = start + format % self.display()
+            print(line, end=' ')
+            try:
+                print(self.method.__doc__)
+            except:
+                print("*** no doc found")
+        else:
+            beg_start = level*' ' + '>>> '
+            end_start = level*' ' + '<<< '
+            trailer = trail * '-'
+            # 4 is the len of '>>> '
+            width = tab - level - 4 - trail
+            format = "%%-%ds" % width
+            beg_line = beg_start + format % self.display() + trail*'>'
+            end_line = end_start + format % self.display() + trail*'<'
+            print(beg_line)
+            for step in self.substeps:
+                Step(step).print_doc(level+1)
+            print(end_line)
+
+    # return a list of (name, method) for all native steps involved
+    def tuples (self):
+        if self.native:
+            return [ (self.internal(), self.method,) ]
+        else:
+            result = []
+            for substep in [ Step(name) for name in self.substeps ] : 
+                result += substep.tuples()
+            return result
+
+    # convenience for listing macros
+    # just do a listdir, hoping we're in the right directory...
+    @staticmethod
+    def list_macros ():
+        names= list(sequences.keys())
+        names.sort()
+        return names
+
+class TestMain:
 
     default_config = [ 'default' ] 
+#    default_rspec_styles = [ 'pl', 'pg' ]
+    default_rspec_styles = [ 'pg' ]
 
-    default_build_url = "http://svn.planet-lab.org/svn/build/trunk"
+    default_build_url = "git://git.onelab.eu/tests"
 
-    def __init__ (self):
-       self.path=os.path.dirname(sys.argv[0]) or "."
+    def __init__(self):
+        self.path = os.path.dirname(sys.argv[0]) or "."
         os.chdir(self.path)
 
-    @staticmethod
-    def show_env (options, message):
-        utils.header (message)
-        utils.show_options("main options",options)
+    def show_env(self, options, message):
+        if self.options.verbose:
+            utils.header(message)
+            utils.show_options("main options", options)
 
-    @staticmethod
-    def optparse_list (option, opt, value, parser):
-        try:
-            setattr(parser.values,option.dest,getattr(parser.values,option.dest)+value.split())
-        except:
-            setattr(parser.values,option.dest,value.split())
+    def init_steps(self):
+        self.steps_message  = ""
+        if not self.options.bonding_build:
+            self.steps_message += 20*'x' + " Defaut steps are\n" + \
+                                  TestPlc.printable_steps(TestPlc.default_steps)
+            self.steps_message += 20*'x' + " Other useful steps are\n" + \
+                                  TestPlc.printable_steps(TestPlc.other_steps)
+            self.steps_message += 20*'x' + " Macro steps are\n" + \
+                                  " ".join(Step.list_macros())
+        else:
+            self.steps_message += 20*'x' + " Default steps with bonding build are\n" + \
+                                  TestPlc.printable_steps(TestPlc.default_bonding_steps)
+
+    def list_steps(self):
+        if not self.options.verbose:
+            print(self.steps_message)
+        else:
+            # steps mentioned on the command line
+            if self.options.steps:
+                scopes = [("Argument steps",self.options.steps)]
+            else:
+                scopes = [("Default steps", TestPlc.default_steps)]
+                if self.options.all_steps:
+                    scopes.append ( ("Other steps", TestPlc.other_steps) )
+                    # try to list macro steps as well
+                    scopes.append ( ("Macro steps", Step.list_macros()) )
+            for (scope, steps) in scopes:
+                print('--------------------', scope)
+                for step in [step for step in steps if TestPlc.valid_step(step)]:
+                    try:
+                        (step, qualifier) = step.split('@')
+                    except:
+                        pass
+                    stepname = step
+                    for special in ['force', 'ignore']:
+                        stepname = stepname.replace('_'+special, "")
+                    Step(stepname).print_doc()
 
     def run (self):
-        steps_message=20*'x'+" Defaut steps are\n"+TestPlc.printable_steps(TestPlc.default_steps)
-        steps_message += "\n"+20*'x'+" Other useful steps are\n"+TestPlc.printable_steps(TestPlc.other_steps)
         usage = """usage: %%prog [options] steps
 arch-rpms-url defaults to the last value used, as stored in arg-arch-rpms-url,
    no default
-build-url defaults to the last value used, as stored in arg-build-url, 
-   or %s
 config defaults to the last value used, as stored in arg-config,
-   or %r
-node-ips and plc-ips defaults to the last value used, as stored in arg-ips-node and arg-ips-plc,
+   or {}
+ips_vnode, ips_vplc and ips_qemu defaults to the last value used, 
+   as stored in arg-ips-{{bplc,vplc,bnode,vnode}},
    default is to use IP scanning
 steps refer to a method in TestPlc or to a step_* module
+
+run with -l to see a list of available steps
 ===
-"""%(TestMain.default_build_url,TestMain.default_config)
-        usage += steps_message
-        parser=OptionParser(usage=usage,version=self.subversion_id)
-        parser.add_option("-u","--url",action="store", dest="arch_rpms_url", 
-                          help="URL of the arch-dependent RPMS area - for locating what to test")
-        parser.add_option("-b","--build",action="store", dest="build_url", 
-                          help="Build URL - for locating vtest-init-vserver.sh")
-        parser.add_option("-c","--config",action="callback", callback=TestMain.optparse_list, dest="config",
-                          nargs=1,type="string",
-                          help="Config module - can be set multiple times, or use quotes")
-        parser.add_option("-p","--personality",action="store", dest="personality", 
-                          help="personality - as in vbuild-nightly")
-        parser.add_option("-d","--pldistro",action="store", dest="pldistro", 
-                          help="pldistro - as in vbuild-nightly")
-        parser.add_option("-f","--fcdistro",action="store", dest="fcdistro", 
-                          help="fcdistro - as in vbuild-nightly")
-        parser.add_option("-x","--exclude",action="callback", callback=TestMain.optparse_list, dest="exclude",
-                          nargs=1,type="string",default=[],
-                          help="steps to exclude - can be set multiple times, or use quotes")
-        parser.add_option("-a","--all",action="store_true",dest="all_steps", default=False,
-                          help="Run all default steps")
-        parser.add_option("-l","--list",action="store_true",dest="list_steps", default=False,
-                          help="List known steps")
-        parser.add_option("-N","--nodes",action="callback", callback=TestMain.optparse_list, dest="ips_node",
-                          nargs=1,type="string",
-                          help="Specify the set of IP addresses to use for nodes (scanning disabled)")
-        parser.add_option("-P","--plcs",action="callback", callback=TestMain.optparse_list, dest="ips_plc",
-                          nargs=1,type="string",
-                          help="Specify the set of IP addresses to use for plcs (scanning disabled)")
-        parser.add_option("-1","--small",action="store_true",dest="small_test",default=False,
-                          help="run a small test -- typically only one node")
-        parser.add_option("-D","--dbname",action="store",dest="dbname",default=None,
-                           help="Used by db_dump and db_restore")
-        parser.add_option("-v","--verbose", action="store_true", dest="verbose", default=False, 
-                          help="Run in verbose mode")
-        parser.add_option("-q","--quiet", action="store_true", dest="quiet", default=False, 
-                          help="Run in quiet mode")
-        parser.add_option("-i","--interactive",action="store_true",dest="interactive",default=False,
-                          help="prompts before each step")
-        parser.add_option("-n","--dry-run", action="store_true", dest="dry_run", default=False,
-                          help="Show environment and exits")
-        parser.add_option("-r","--restart-nm", action="store_true", dest="forcenm", default=False, 
-                          help="Force the NM to restart in check_slices step")
-        parser.add_option("-t","--trace", action="store", dest="trace_file", default=None,
-                          #default="logs/trace-@TIME@.txt",
-                          help="Trace file location")
-        (self.options, self.args) = parser.parse_args()
-
-        if len(self.args) == 0:
-            if self.options.all_steps:
-                self.options.steps=TestPlc.default_steps
-            elif self.options.dry_run:
-                self.options.steps=TestPlc.default_steps
-            elif self.options.list_steps:
-                print steps_message
+""".format(TestMain.default_config)
+
+        parser = ArgumentParser(usage = usage)
+        parser.add_argument("-u", "--url", action="store",  dest="arch_rpms_url", 
+                            help="URL of the arch-dependent RPMS area - for locating what to test")
+        parser.add_argument("-b", "--build", action="store", dest="build_url", 
+                            help="ignored, for legacy only")
+        parser.add_argument("-c", "--config", action="append", dest="config", default=[],
+                            help="Config module - can be set multiple times, or use quotes")
+        parser.add_argument("-p", "--personality", action="store", dest="personality", 
+                            help="personality - as in vbuild-nightly")
+        parser.add_argument("-d", "--pldistro", action="store", dest="pldistro", 
+                            help="pldistro - as in vbuild-nightly")
+        parser.add_argument("-f", "--fcdistro", action="store", dest="fcdistro", 
+                            help="fcdistro - as in vbuild-nightly")
+        parser.add_argument("-e", "--exclude", action="append", dest="exclude", default=[],
+                            help="steps to exclude - can be set multiple times, or use quotes")
+        parser.add_argument("-i", "--ignore", action="append", dest="ignore", default=[],
+                            help="steps to run but ignore - can be set multiple times, or use quotes")
+        parser.add_argument("-a", "--all", action="store_true", dest="all_steps", default=False,
+                            help="Run all default steps")
+        parser.add_argument("-l", "--list", action="store_true", dest="list_steps", default=False,
+                            help="List known steps")
+        parser.add_argument("-V", "--vserver", action="append", dest="ips_bplc", default=[],
+                            help="Specify the set of hostnames for the boxes that host the plcs")
+        parser.add_argument("-P", "--plcs", action="append", dest="ips_vplc", default=[],
+                            help="Specify the set of hostname/IP's to use for vplcs")
+        parser.add_argument("-Q", "--qemus", action="append", dest="ips_bnode", default=[],
+                            help="Specify the set of hostnames for the boxes that host the nodes")
+        parser.add_argument("-N", "--nodes", action="append", dest="ips_vnode", default=[],
+                            help="Specify the set of hostname/IP's to use for vnodes")
+        parser.add_argument("-s", "--size", action="store", dest="size", default=1,
+                            type=int, 
+                            help="set test size in # of plcs - default is 1")
+        parser.add_argument("-q", "--qualifier", action="store", dest="qualifier", default=None,
+                            type=int, 
+                            help="run steps only on plc numbered <qualifier>, starting at 1")
+        parser.add_argument("-y", "--rspec-style", action="append", dest="rspec_styles", default=[],
+                            help="pl is for planetlab rspecs, pg is for protogeni")
+        parser.add_argument("-k", "--keep-going", action="store", dest="keep_going", default=False,
+                            help="proceeds even if some steps are failing")
+        parser.add_argument("-D", "--dbname", action="store", dest="dbname", default=None,
+                            help="Used by plc_db_dump and plc_db_restore")
+        parser.add_argument("-v", "--verbose", action="store_true", dest="verbose", default=False, 
+                            help="Run in verbose mode")
+        parser.add_argument("-I", "--interactive", action="store_true", dest="interactive", default=False,
+                            help="prompts before each step")
+        parser.add_argument("-n", "--dry-run", action="store_true", dest="dry_run", default=False,
+                            help="Show environment and exits")
+        parser.add_argument("-t", "--trace", action="store", dest="trace_file", default=None,
+                            help="Trace file location")
+        parser.add_argument("-g", "--bonding", action='store', dest='bonding_build', default=None,
+                            help="specify build to bond with")
+        # if we call symlink 'rung' instead of just run this is equivalent to run -G
+        bonding_default = 'rung' in sys.argv[0]
+        parser.add_argument("-G", "--bonding-env", action='store_true', dest='bonding_env', default=bonding_default,
+                            help="get bonding build from env. variable $bonding")
+        parser.add_argument("steps", nargs='*')
+        self.options = parser.parse_args()
+
+        # handle -G/-g options
+        if self.options.bonding_env:
+            if 'bonding' not in os.environ:
+                print("env. variable $bonding must be set with --bonding-env")
                 sys.exit(1)
-            else:
-                print 'No step found (do you mean -a ? )'
-                print "Run %s --help for help"%sys.argv[0]                        
+            self.options.bonding_build = os.environ['bonding']
+
+        if self.options.bonding_build:
+            ## allow to pass -g ../2015.03.15--f18 so we can use bash completion
+            self.options.bonding_build = os.path.basename(self.options.bonding_build)
+            if not os.path.isdir("../{}".format(self.options.bonding_build)):
+                print("could not find test dir for bonding build {}".format(self.options.bonding_build))
                 sys.exit(1)
-        else:
-            self.options.steps = self.args
+
+        # allow things like "run -c 'c1 c2' -c c3"
+        def flatten (x):
+            result = []
+            for el in x:
+                if hasattr(el, "__iter__") and not isinstance(el, str):
+                    result.extend(flatten(el))
+                else:
+                    result.append(el)
+            return result
+        # flatten relevant options
+        for optname in ['config', 'exclude', 'ignore', 'ips_bplc', 'ips_vplc', 'ips_bnode', 'ips_vnode']:
+            setattr(self.options, optname,
+                    flatten([arg.split() for arg in getattr(self.options, optname)]))
+
+        if not self.options.rspec_styles:
+            self.options.rspec_styles = TestMain.default_rspec_styles
 
         # handle defaults and option persistence
-        for (recname,filename,default) in (
-            ('build_url','arg-build-url',TestMain.default_build_url) ,
-            ('ips_node','arg-ips-node',[]) , 
-            ('ips_plc','arg-ips-plc',[]) , 
-            ('config','arg-config',TestMain.default_config) , 
-            ('arch_rpms_url','arg-arch-rpms-url',"") , 
-            ('personality','arg-personality',"linux32"),
-            ('pldistro','arg-pldistro',"planetlab"),
-            ('fcdistro','arg-fcdistro','f8'),
-            ) :
-#            print 'handling',recname
-            path=filename
-            is_list = isinstance(default,list)
-            if not getattr(self.options,recname):
+        for recname, filename, default, need_reverse in (
+            ('build_url', 'arg-build-url', TestMain.default_build_url, None),
+            ('ips_bplc', 'arg-ips-bplc', [], True),
+            ('ips_vplc', 'arg-ips-vplc', [], True), 
+            ('ips_bnode', 'arg-ips-bnode', [], True),
+            ('ips_vnode', 'arg-ips-vnode', [], True), 
+            ('config', 'arg-config', TestMain.default_config, False), 
+            ('arch_rpms_url', 'arg-arch-rpms-url', "", None), 
+            ('personality', 'arg-personality', "linux64", None),
+            ('pldistro', 'arg-pldistro', "onelab", None),
+            ('fcdistro', 'arg-fcdistro', 'f14', None),
+            ):
+#            print('handling', recname)
+            path = filename
+            is_list = isinstance(default, list)
+            is_bool = isinstance(default, bool)
+            if not getattr(self.options, recname):
                 try:
-                    parsed=file(path).readlines()
-                    if not is_list:    # strings
+                    with open(path) as file:
+                        parsed = file.readlines()
+                    if is_list:         # lists
+                        parsed = [x.strip() for x in parsed]
+                    else:               # strings and booleans
                         if len(parsed) != 1:
-                            print "%s - error when parsing %s"%(sys.argv[1],path)
+                            print("{} - error when parsing {}".format(sys.argv[1], path))
                             sys.exit(1)
-                        parsed=parsed[0].strip()
-                    else:              # lists
-                        parsed=[x.strip() for x in parsed]
-                    setattr(self.options,recname,parsed)
-                except:
+                        parsed = parsed[0].strip()
+                        if is_bool:
+                            parsed = parsed.lower() == 'true'
+                    setattr(self.options, recname, parsed)
+                except  Exception as e:
                     if default != "":
-                        setattr(self.options,recname,default)
+                        setattr(self.options, recname, default)
                     else:
-                        print "Cannot determine",recname
-                        print "Run %s --help for help"%sys.argv[0]                        
+                        print("Cannot determine", recname, e)
+                        print("Run {} --help for help".format(sys.argv[0]))
                         sys.exit(1)
-            if not self.options.quiet:
-                utils.header('* Using %s = %s'%(recname,getattr(self.options,recname)))
 
             # save for next run
-            fsave=open(path,"w")
-            if not is_list:
-                fsave.write(getattr(self.options,recname) + "\n")
-            else:
-                for value in getattr(self.options,recname):
+            fsave = open(path, "w")
+            if is_list:                 # lists
+                for value in getattr(self.options, recname):
                     fsave.write(value + "\n")
+            else:                       # strings and booleans - just call str()
+                fsave.write(str(getattr(self.options, recname)) + "\n")
             fsave.close()
-#            utils.header('Saved %s into %s'%(recname,filename))
+#            utils.header('Saved {} into {}'.format(recname, filename))
 
-        if self.options.personality == "linux32":
-            self.options.arch = "i386"
-        elif self.options.personality == "linux64":
-            self.options.arch = "x86_64"
-        else:
-            raise Exception, "Unsupported personality %r"%self.options.personality
-        # steps
+            # lists need be reversed
+            # I suspect this is useful for the various pools but for config, it's painful
+            if isinstance(getattr(self.options, recname), list) and need_reverse:
+                getattr(self.options, recname).reverse()
+
+            if self.options.verbose:
+                utils.header('* Using {} = {}'.format(recname, getattr(self.options, recname)))
+
+        # hack : if sfa is not among the published rpms, skip these tests
+        TestPlc.check_whether_build_has_sfa(self.options.arch_rpms_url)
+
+        # initialize steps
         if not self.options.steps:
-            #default (all) steps
-            #self.options.steps=['dump','clean','install','populate']
-            self.options.steps=TestPlc.default_steps
+            # defaults, depends on using bonding or not
+            if self.options.bonding_build:
+                self.options.steps = TestPlc.default_bonding_steps
+            else:
+                self.options.steps = TestPlc.default_steps
+
+        if self.options.list_steps:
+            self.init_steps()
+            self.list_steps()
+            return 'SUCCESS'
+
+        # rewrite '-' into '_' in step names
+        self.options.steps   = [ step.replace('-', '_') for step in self.options.steps ]
+        self.options.exclude = [ step.replace('-', '_') for step in self.options.exclude ]
+        self.options.ignore  = [ step.replace('-', '_') for step in self.options.ignore ]
+
+        # technicality, decorate known steps to produce the '_ignore' version
+        TestPlc.create_ignore_steps()
 
         # exclude
-        selected=[]
+        selected = []
         for step in self.options.steps:
-            keep=True
+            keep = True
             for exclude in self.options.exclude:
-                if utils.match(step,exclude):
-                    keep=False
+                if utils.match(step, exclude):
+                    keep = False
                     break
-            if keep: selected.append(step)
-        self.options.steps=selected
+            if keep:
+                selected.append(step)
+
+        # ignore
+        selected = [ step if step not in self.options.ignore else step + "_ignore"
+                     for step in selected ]
+
+        self.options.steps = selected
 
         # this is useful when propagating on host boxes, to avoid conflicts
-        self.options.buildname = os.path.basename (os.path.abspath (self.path))
+        self.options.buildname = os.path.basename(os.path.abspath(self.path))
 
         if self.options.verbose:
-            self.show_env(self.options,"Verbose")
+            self.show_env(self.options, "Verbose")
 
         # load configs
         all_plc_specs = []
         for config in self.options.config:
-            modulename='config_'+config
+            modulename = 'config_' + config
             try:
                 m = __import__(modulename)
-                all_plc_specs = m.config(all_plc_specs,self.options)
+                all_plc_specs = m.config(all_plc_specs, self.options)
             except :
                 traceback.print_exc()
-                print 'Cannot load config %s -- ignored'%modulename
+                print('Cannot load config {} -- ignored'.format(modulename))
                 raise
-        # remember plc IP address(es) if not specified
-        current=file('arg-ips-plc').read()
-        if not current:
-            ips_plc_file=open('arg-ips-plc','w')
+
+        # provision on local substrate
+        all_plc_specs = LocalSubstrate.local_substrate.provision(all_plc_specs, self.options)
+
+        # remember substrate IP address(es) for next run
+        with open('arg-ips-bplc', 'w') as ips_bplc_file:
             for plc_spec in all_plc_specs:
-                ips_plc_file.write("%s\n"%plc_spec['PLC_API_HOST'])
-            ips_plc_file.close()
+                ips_bplc_file.write("{}\n".format(plc_spec['host_box']))
+        with open('arg-ips-vplc', 'w') as ips_vplc_file:
+            for plc_spec in all_plc_specs:
+                ips_vplc_file.write("{}\n".format(plc_spec['settings']['PLC_API_HOST']))
         # ditto for nodes
-        current=file('arg-ips-node').read()
-        if not current:
-            ips_node_file=open('arg-ips-node','w')
+        with open('arg-ips-bnode', 'w') as ips_bnode_file:
+            for plc_spec in all_plc_specs:
+                for site_spec in plc_spec['sites']:
+                    for node_spec in site_spec['nodes']:
+                        ips_bnode_file.write("{}\n".format(node_spec['host_box']))
+        with open('arg-ips-vnode','w') as ips_vnode_file:
             for plc_spec in all_plc_specs:
                 for site_spec in plc_spec['sites']:
                     for node_spec in site_spec['nodes']:
-                        ips_node_file.write("%s\n"%node_spec['node_fields']['hostname'])
-            ips_node_file.close()
+                        # back to normal (unqualified) form
+                        stripped = node_spec['node_fields']['hostname'].split('.')[0]
+                        ips_vnode_file.write("{}\n".format(stripped))
+
         # build a TestPlc object from the result, passing options
         for spec in all_plc_specs:
-            spec['disabled'] = False
+            spec['failed_step'] = False
         all_plcs = [ (x, TestPlc(x,self.options)) for x in all_plc_specs]
 
         # pass options to utils as well
         utils.init_options(self.options)
 
-        overall_result = True
-        testplc_method_dict = __import__("TestPlc").__dict__['TestPlc'].__dict__
-        all_step_infos=[]
+        # populate TestBonding objects
+        # need to wait until here as we need all_plcs
+        if self.options.bonding_build:
+            # this will fail if ../{bonding_build} has not the right arg- files
+            for spec, test_plc in all_plcs:
+                test_plc.test_bonding = TestBonding (test_plc,
+                                                     onelab_bonding_spec(self.options.bonding_build),
+                                                     LocalSubstrate.local_substrate,
+                                                     self.options)
+        
+        overall_result = 'SUCCESS'
+        all_step_infos = []
         for step in self.options.steps:
             if not TestPlc.valid_step(step):
                 continue
-            force=False
-            # is it a forced step
-            if step.find("force_") == 0:
-                step=step.replace("force_","")
-                force=True
-            # try and locate a method in TestPlc
-            if testplc_method_dict.has_key(step):
-                all_step_infos += [ (step, testplc_method_dict[step] , force)]
-            # otherwise search for the 'run' method in the step_<x> module
-            else:
-                modulename='step_'+step
-                try:
-                    # locate all methods named run* in the module
-                    module_dict = __import__(modulename).__dict__
-                    names = [ key for key in module_dict.keys() if key.find("run")==0 ]
-                    if not names:
-                        raise Exception,"No run* method in module %s"%modulename
-                    names.sort()
-                    all_step_infos += [ ("%s.%s"%(step,name),module_dict[name],force) for name in names ]
-                except :
-                    print '********** step %s NOT FOUND -- ignored'%(step)
-                    traceback.print_exc()
-                    overall_result = False
+            # some steps need to be done regardless of the previous ones: we force them
+            force = False
+            if step.endswith("_force"):
+                step = step.replace("_force", "")
+                force = True
+            # allow for steps to specify an index like in 
+            # run checkslice@2
+            try:
+                step, qualifier = step.split('@')
+            except:
+                qualifier = self.options.qualifier
+
+            try:
+                stepobj = Step (step)
+                for substep, method in stepobj.tuples():
+                    # a cross step will run a method on TestPlc that has a signature like
+                    # def cross_foo (self, all_test_plcs)
+                    cross = False
+                    if substep.find("cross_") == 0:
+                        cross = True
+                    all_step_infos.append ( (substep, method, force, cross, qualifier, ) )
+            except :
+                utils.header("********** FAILED step {} (NOT FOUND) -- won't be run".format(step))
+                traceback.print_exc()
+                overall_result = 'FAILURE'
             
         if self.options.dry_run:
-            self.show_env(self.options,"Dry run")
+            self.show_env(self.options, "Dry run")
         
         # init & open trace file if provided
         if self.options.trace_file and not self.options.dry_run:
-            time=strftime("%H-%M")
-            date=strftime("%Y-%m-%d")
-            trace_file=self.options.trace_file
-            trace_file=trace_file.replace("@TIME@",time)
-            trace_file=trace_file.replace("@DATE@",date)
-            self.options.trace_file=trace_file
             # create dir if needed
-            trace_dir=os.path.dirname(trace_file)
+            trace_dir = os.path.dirname(self.options.trace_file)
             if trace_dir and not os.path.isdir(trace_dir):
                 os.makedirs(trace_dir)
-            trace=open(trace_file,"w")
+            trace = open(self.options.trace_file,"w")
 
         # do all steps on all plcs
-        TRACE_FORMAT="TRACE: time=%(time)s plc=%(plcname)s step=%(stepname)s status=%(status)s force=%(force)s\n"
-        for (stepname,method,force) in all_step_infos:
-            for (spec,obj) in all_plcs:
-                plcname=spec['name']
+        TIME_FORMAT = "%H-%M-%S"
+        TRACE_FORMAT = "TRACE: {plc_counter:d} {begin}->{seconds}s={duration}s " + \
+                       "status={status} step={stepname} plc={plcname} force={force}\n"
+        for stepname, method, force, cross, qualifier in all_step_infos:
+            plc_counter = 0
+            for spec, plc_obj in all_plcs:
+                plc_counter += 1
+                # skip this step if we have specified a plc_explicit
+                if qualifier and plc_counter != int(qualifier):
+                    continue
+
+                plcname = spec['name']
+                across_plcs = [ o for (s,o) in all_plcs if o!=plc_obj ]
 
                 # run the step
-                time=strftime("%Y-%m-%d-%H-%M")
-                if not spec['disabled'] or force or self.options.interactive:
+                beg_time = datetime.now()
+                begin = beg_time.strftime(TIME_FORMAT)
+                if not spec['failed_step'] or force or self.options.interactive or self.options.keep_going:
+                    skip_step = False
                     if self.options.interactive:
-                        msg="Run step %s on %s [y]/n/q ? "%(stepname,plcname)
-                        answer=raw_input(msg).strip().lower() or "y"
-                        answer=answer[0]
-                        if answer in ['n']:
-                            print '%s on %s skipped'%(stepname,plcname)
-                            continue
-                        elif answer in ['q','b']:
-                            print 'Exiting'
-                            return
+                        prompting = True
+                        while prompting:
+                            msg="{:d} Run step {} on {} [r](un)/d(ry_run)/p(roceed)/s(kip)/q(uit) ? "\
+                                .format(plc_counter, stepname, plcname)
+                            answer = input(msg).strip().lower() or "r"
+                            answer = answer[0]
+                            if answer in ['s','n']:     # skip/no/next
+                                print('{} on {} skipped'.format(stepname, plcname))
+                                prompting = False
+                                skip_step = True
+                            elif answer in ['q','b']:   # quit/bye
+                                print('Exiting')
+                                return 'FAILURE'
+                            elif answer in ['d']:       # dry_run
+                                dry_run = self.options.dry_run
+                                self.options.dry_run = True
+                                plc_obj.options.dry_run = True
+                                plc_obj.apiserver.set_dry_run(True)
+                                if not cross:
+                                    step_result=method(plc_obj)
+                                else:
+                                    step_result=method(plc_obj, across_plcs)
+                                print('dry_run step ->', step_result)
+                                self.options.dry_run = dry_run
+                                plc_obj.options.dry_run = dry_run
+                                plc_obj.apiserver.set_dry_run(dry_run)
+                            elif answer in ['p']:
+                                # take it as a yes and leave interactive mode
+                                prompting = False
+                                self.options.interactive = False
+                            elif answer in ['r','y']:   # run/yes
+                                prompting = False
+                    if skip_step:
+                        continue
                     try:
-                        force_msg=""
-                        if force: force_msg=" (forced)"
-                        utils.header("********** RUNNING step %s%s on plc %s"%(stepname,force_msg,plcname))
-                        step_result = method(obj)
-                        if step_result:
-                            utils.header('********** SUCCESSFUL step %s on %s'%(stepname,plcname))
-                            status="OK"
+                        force_msg = ""
+                        if force and spec['failed_step']:
+                            force_msg=" (forced after {} has failed)".format(spec['failed_step'])
+                        utils.header("********** {:d} RUNNING step {}{} on plc {}"\
+                                     .format(plc_counter, stepname, force_msg, plcname))
+                        if not cross:
+                            step_result = method(plc_obj)
                         else:
-                            overall_result = False
-                            spec['disabled'] = True
-                            utils.header('********** FAILED Step %s on %s - discarding that plc from further steps'%(stepname,plcname))
-                            status="KO"
+                            step_result = method(plc_obj, across_plcs)
+                        if isinstance (step_result, Ignored):
+                            step_result = step_result.result
+                            if step_result:
+                                msg = "OK"
+                            else:
+                                msg = "KO"
+                                # do not overwrite if FAILURE
+                                if overall_result == 'SUCCESS': 
+                                    overall_result = 'IGNORED'
+                            utils.header('********** {} IGNORED ({}) step {} on {}'\
+                                         .format(plc_counter, msg, stepname, plcname))
+                            status="{}[I]".format(msg)
+                        elif step_result:
+                            utils.header('********** {:d} SUCCESSFUL step {} on {}'\
+                                         .format(plc_counter, stepname, plcname))
+                            status = "OK"
+                        else:
+                            overall_result = 'FAILURE'
+                            spec['failed_step'] = stepname
+                            utils.header('********** {:d} FAILED step {} on {} (discarded from further steps)'\
+                                         .format(plc_counter, stepname, plcname))
+                            status = "KO"
                     except:
-                        overall_result=False
-                        spec['disabled'] = True
+                        overall_result = 'FAILURE'
+                        spec['failed_step'] = stepname
                         traceback.print_exc()
-                        utils.header ('********** FAILED (exception) Step %s on plc %s - discarding this plc from further steps'%(stepname,plcname))
-                        status="KO"
+                        utils.header ('********** {} FAILED (exception) step {} on {} (discarded from further steps)'\
+                                      .format(plc_counter, stepname, plcname))
+                        status = "KO"
 
                 # do not run, just display it's skipped
                 else:
-                    utils.header("********** IGNORED Plc %s is disabled - skipping step %s"%(plcname,stepname))
-                    status="UNDEF"
+                    why = "has failed {}".format(spec['failed_step'])
+                    utils.header("********** {} SKIPPED Step {} on {} ({})"\
+                                 .format(plc_counter, stepname, plcname, why))
+                    status = "UNDEF"
                 if not self.options.dry_run:
-                    # alwas do this on stdout
-                    print TRACE_FORMAT%locals()
+                    delay = datetime.now()-beg_time
+                    seconds = int(delay.total_seconds())
+                    duration = str(delay)
+                    # always do this on stdout
+                    print(TRACE_FORMAT.format(**locals()))
                     # duplicate on trace_file if provided
                     if self.options.trace_file:
-                        trace.write(TRACE_FORMAT%locals())
+                        trace.write(TRACE_FORMAT.format(**locals()))
+                        trace.flush()
 
         if self.options.trace_file and not self.options.dry_run:
             trace.close()
 
+        # free local substrate
+        LocalSubstrate.local_substrate.release(self.options)
+        
         return overall_result
 
     # wrapper to run, returns a shell-compatible result
+    # retcod:
+    # 0: SUCCESS
+    # 1: FAILURE
+    # 2: SUCCESS but some ignored steps failed
+    # 3: OTHER ERROR
     def main(self):
         try:
-            success=self.run()
-            if success:
+            success = self.run()
+            if success == 'SUCCESS':
                 return 0
+            elif success == 'IGNORED':
+                return 2
             else:
-                return 1 
+                return 1
         except SystemExit:
-            raise
+            print('Caught SystemExit')
+            return 3
         except:
             traceback.print_exc()
-            return 2
+            return 3
 
 if __name__ == "__main__":
-    sys.exit(TestMain().main())
+    exit_code = TestMain().main()
+    print("TestMain exit code", exit_code)
+    sys.exit(exit_code)