def v4_graph():
from lnt.server.ui import util
from lnt.testing import PASS
from lnt.util import stats
from lnt.external.stats import stats as ext_stats
ts = request.get_testsuite()
# Parse the view options.
options = {}
options['hide_lineplot'] = bool(request.args.get('hide_lineplot'))
show_lineplot = not options['hide_lineplot']
options['show_mad'] = show_mad = bool(request.args.get('show_mad'))
options['show_stddev'] = show_stddev = bool(request.args.get('show_stddev'))
options['hide_all_points'] = hide_all_points = bool(
request.args.get('hide_all_points'))
options['show_linear_regression'] = show_linear_regression = bool(
request.args.get('show_linear_regression'))
options['show_failures'] = show_failures = bool(
request.args.get('show_failures'))
options['normalize_by_median'] = normalize_by_median = bool(
request.args.get('normalize_by_median'))
options['show_moving_average'] = moving_average = bool(
request.args.get('show_moving_average'))
options['show_moving_median'] = moving_median = bool(
request.args.get('show_moving_median'))
options['moving_window_size'] = moving_window_size = int(
request.args.get('moving_window_size', 10))
options['hide_highlight'] = bool(
request.args.get('hide_highlight'))
show_highlight = not options['hide_highlight']
def convert_revision(dotted):
"""Turn a version number like 489.2.10 into something
that is ordered and sortable.
For now 489.2.10 will be returned as a tuple of ints.
"""
dotted = integral_rex.findall(dotted)
return tuple([int(d) for d in dotted])
# Load the graph parameters.
graph_parameters = []
for name,value in request.args.items():
# Plots to graph are passed as::
#
# plot.<unused>=<machine id>.<test id>.<field index>
if not name.startswith(str('plot.')):
continue
# Ignore the extra part of the key, it is unused.
machine_id_str,test_id_str,field_index_str = value.split('.')
try:
machine_id = int(machine_id_str)
test_id = int(test_id_str)
field_index = int(field_index_str)
except:
return abort(400)
if not (0 <= field_index < len(ts.sample_fields)):
return abort(404)
try:
machine = \
ts.query(ts.Machine).filter(ts.Machine.id == machine_id).one()
test = ts.query(ts.Test).filter(ts.Test.id == test_id).one()
field = ts.sample_fields[field_index]
except NoResultFound:
return abort(404)
graph_parameters.append((machine, test, field))
# Order the plots by machine name, test name and then field.
graph_parameters.sort(key = lambda (m,t,f): (m.name, t.name, f.name))
# Extract requested mean trend.
mean_parameter = None
for name,value in request.args.items():
# Mean to graph is passed as:
#
# mean=<machine id>.<field index>
if name != 'mean':
continue
machine_id_str,field_index_str = value.split('.')
try:
machine_id = int(machine_id_str)
field_index = int(field_index_str)
except ValueError:
return abort(400)
if not (0 <= field_index < len(ts.sample_fields)):
return abort(404)
try:
machine = \
ts.query(ts.Machine).filter(ts.Machine.id == machine_id).one()
except NoResultFound:
return abort(404)
field = ts.sample_fields[field_index]
mean_parameter = (machine, field)
# Sanity check the arguments.
if not graph_parameters and not mean_parameter:
return render_template("error.html", message="Nothing to graph.")
# Extract requested baselines, and their titles.
baseline_parameters = []
for name,value in request.args.items():
# Baselines to graph are passed as:
#
# baseline.title=<run id>
if not name.startswith(str('baseline.')):
continue
baseline_title = name[len('baseline.'):]
run_id_str = value
try:
run_id = int(run_id_str)
except:
return abort(400)
try:
run = ts.query(ts.Run).join(ts.Machine).filter(ts.Run.id == run_id).one()
except:
err_msg = "The run {} was not found in the database.".format(run_id)
return render_template("error.html",
message=err_msg)
baseline_parameters.append((run, baseline_title))
# Create region of interest for run data region if we are performing a
# comparison.
revision_range = None
highlight_run_id = request.args.get('highlight_run')
if show_highlight and highlight_run_id and highlight_run_id.isdigit():
highlight_run = ts.query(ts.Run).filter_by(
id=int(highlight_run_id)).first()
if highlight_run is None:
abort(404)
# Find the neighboring runs, by order.
prev_runs = list(ts.get_previous_runs_on_machine(highlight_run, N = 1))
if prev_runs:
start_rev = prev_runs[0].order.llvm_project_revision
end_rev = highlight_run.order.llvm_project_revision
revision_range = {
"start": convert_revision(start_rev),
"end": convert_revision(end_rev) }
# Build the graph data.
legend = []
graph_plots = []
graph_datum = []
overview_plots = []
baseline_plots = []
num_plots = len(graph_parameters)
for i,(machine,test,field) in enumerate(graph_parameters):
# Determine the base plot color.
col = list(util.makeDarkColor(float(i) / num_plots))
legend.append((machine, test.name, field.name, tuple(col)))
# Load all the field values for this test on the same machine.
#
# FIXME: Don't join to Order here, aggregate this across all the tests
# we want to load. Actually, we should just make this a single query.
#
# FIXME: Don't hard code field name.
q = ts.query(field.column, ts.Order.llvm_project_revision, ts.Run.start_time).\
join(ts.Run).join(ts.Order).\
filter(ts.Run.machine_id == machine.id).\
filter(ts.Sample.test == test).\
filter(field.column != None)
# Unless all samples requested, filter out failing tests.
if not show_failures:
if field.status_field:
q = q.filter((field.status_field.column == PASS) |
(field.status_field.column == None))
# Aggregate by revision.
data = util.multidict((rev, (val, date)) for val,rev,date in q).items()
data.sort(key=lambda sample: convert_revision(sample[0]))
graph_datum.append((test.name, data, col, field))
# Get baselines for this line
num_baselines = len(baseline_parameters)
for baseline_id, (baseline, baseline_title) in enumerate(baseline_parameters):
q_baseline = ts.query(field.column, ts.Order.llvm_project_revision, ts.Run.start_time, ts.Machine.name).\
join(ts.Run).join(ts.Order).join(ts.Machine).\
filter(ts.Run.id == baseline.id).\
filter(ts.Sample.test == test).\
filter(field.column != None)
# In the event of many samples, use the mean of the samples as the baseline.
samples = []
for sample in q_baseline:
samples.append(sample[0])
# Skip this baseline if there is no data.
if not samples:
continue
mean = sum(samples)/len(samples)
# Darken the baseline color distinguish from non-baselines.
# Make a color closer to the sample than its neighbour.
color_offset = float(baseline_id) / num_baselines / 2
my_color = (i + color_offset) / num_plots
dark_col = list(util.makeDarkerColor(my_color))
str_dark_col = util.toColorString(dark_col)
baseline_plots.append({'color': str_dark_col,
'lineWidth': 2,
'yaxis': {'from': mean, 'to': mean},
'name': q_baseline[0].llvm_project_revision})
baseline_name = "Baseline {} on {}".format(baseline_title, q_baseline[0].name)
legend.append((BaselineLegendItem(baseline_name, baseline.id), test.name, field.name, dark_col))
# Draw mean trend if requested.
if mean_parameter:
machine, field = mean_parameter
test_name = 'Geometric Mean'
col = (0,0,0)
legend.append((machine, test_name, field.name, col))
q = ts.query(sqlalchemy.sql.func.min(field.column),
ts.Order.llvm_project_revision,
sqlalchemy.sql.func.min(ts.Run.start_time)).\
join(ts.Run).join(ts.Order).join(ts.Test).\
filter(ts.Run.machine_id == machine.id).\
filter(field.column != None).\
group_by(ts.Order.llvm_project_revision, ts.Test)
# Calculate geomean of each revision.
data = util.multidict(((rev, date), val) for val,rev,date in q).items()
data = [(rev, [(lnt.server.reporting.analysis.calc_geomean(vals), date)])
for ((rev, date), vals) in data]
# Sort data points according to revision number.
data.sort(key=lambda sample: convert_revision(sample[0]))
graph_datum.append((test_name, data, col, field))
for name, data, col, field in graph_datum:
# Compute the graph points.
errorbar_data = []
points_data = []
pts = []
moving_median_data = []
moving_average_data = []
if normalize_by_median:
normalize_by = 1.0/stats.median([min([d[0] for d in values])
for _,values in data])
else:
normalize_by = 1.0
for pos, (point_label, datapoints) in enumerate(data):
# Get the samples.
data = [data_date[0] for data_date in datapoints]
# And the date on which they were taken.
dates = [data_date[1] for data_date in datapoints]
# When we can, map x-axis to revisions, but when that is too hard
# use the position of the sample instead.
rev_x = convert_revision(point_label)
x = rev_x[0] if len(rev_x)==1 else pos
values = [v*normalize_by for v in data]
aggregation_fn = min
if field.bigger_is_better:
aggregation_fn = max
agg_value, agg_index = \
aggregation_fn((value, index)
for (index, value) in enumerate(values))
# Generate metadata.
metadata = {"label":point_label}
metadata["date"] = str(dates[agg_index])
if len(graph_datum) > 1:
# If there are more than one plot in the graph, also label the
# test name.
metadata["test_name"] = name
pts.append((x, agg_value, metadata))
# Add the individual points, if requested.
# For each point add a text label for the mouse over.
if not hide_all_points:
for i,v in enumerate(values):
point_metadata = dict(metadata)
point_metadata["date"] = str(dates[i])
points_data.append((x, v, point_metadata))
# Add the standard deviation error bar, if requested.
if show_stddev:
mean = stats.mean(values)
sigma = stats.standard_deviation(values)
errorbar_data.append((x, mean, sigma))
# Add the MAD error bar, if requested.
if show_mad:
med = stats.median(values)
mad = stats.median_absolute_deviation(values, med)
errorbar_data.append((x, med, mad))
# Compute the moving average and or moving median of our data if requested.
if moving_average or moving_median:
fun = None
def compute_moving_average(x, window, average_list, median_list):
average_list.append((x, lnt.util.stats.mean(window)))
def compute_moving_median(x, window, average_list, median_list):
median_list.append((x, lnt.util.stats.median(window)))
def compute_moving_average_and_median(x, window, average_list, median_list):
average_list.append((x, lnt.util.stats.mean(window)))
median_list.append((x, lnt.util.stats.median(window)))
if moving_average and moving_median:
fun = compute_moving_average_and_median
elif moving_average:
fun = compute_moving_average
else:
fun = compute_moving_median
len_pts = len(pts)
for i in range(len_pts):
start_index = max(0, i - moving_window_size)
end_index = min(len_pts, i + moving_window_size)
window_pts = [x[1] for x in pts[start_index:end_index]]
fun(pts[i][0], window_pts, moving_average_data, moving_median_data)
# On the overview, we always show the line plot.
overview_plots.append({
"data" : pts,
"color" : util.toColorString(col) })
# Add the minimum line plot, if requested.
if show_lineplot:
graph_plots.append({
"data" : pts,
"color" : util.toColorString(col) })
# Add regression line, if requested.
if show_linear_regression:
xs = [t for t,v,_ in pts]
ys = [v for t,v,_ in pts]
# We compute the regression line in terms of a normalized X scale.
x_min, x_max = min(xs), max(xs)
try:
norm_xs = [(x - x_min) / (x_max - x_min)
for x in xs]
except ZeroDivisionError:
norm_xs = xs
try:
info = ext_stats.linregress(norm_xs, ys)
except ZeroDivisionError:
info = None
except ValueError:
info = None
if info is not None:
slope, intercept,_,_,_ = info
reglin_col = [c * .7 for c in col]
reglin_pts = [(x_min, 0.0 * slope + intercept),
(x_max, 1.0 * slope + intercept)]
graph_plots.insert(0, {
"data" : reglin_pts,
"color" : util.toColorString(reglin_col),
"lines" : {
"lineWidth" : 2 },
"shadowSize" : 4 })
# Add the points plot, if used.
if points_data:
pts_col = (0,0,0)
graph_plots.append({
"data" : points_data,
"color" : util.toColorString(pts_col),
"lines" : {
"show" : False },
"points" : {
"show" : True,
"radius" : .25,
"fill" : True } })
# Add the error bar plot, if used.
if errorbar_data:
bar_col = [c*.7 for c in col]
graph_plots.append({
"data" : errorbar_data,
"lines" : { "show" : False },
"color" : util.toColorString(bar_col),
"points" : {
"errorbars" : "y",
"yerr" : { "show" : True,
"lowerCap" : "-",
"upperCap" : "-",
"lineWidth" : 1 } } })
# Add the moving average plot, if used.
if moving_average_data:
col = [0.32, 0.6, 0.0]
graph_plots.append({
"data" : moving_average_data,
"color" : util.toColorString(col) })
# Add the moving median plot, if used.
if moving_median_data:
col = [0.75, 0.0, 1.0]
graph_plots.append({
"data" : moving_median_data,
"color" : util.toColorString(col) })
if bool(request.args.get('json')):
json_obj = dict()
json_obj['data'] = graph_plots
# Flatten ORM machine objects to their string names.
simple_type_legend = []
for machine, test, unit, color in legend:
# Flatten name, make color a dict.
new_entry = {'name': machine.name,
'test': test,
'unit': unit,
'color': util.toColorString(color),}
simple_type_legend.append(new_entry)
json_obj['legend'] = simple_type_legend
json_obj['revision_range'] = revision_range
json_obj['current_options'] = options
json_obj['test_suite_name'] = ts.name
json_obj['baselines'] = baseline_plots
return flask.jsonify(**json_obj)
return render_template("v4_graph.html", ts=ts, options=options,
revision_range=revision_range,
graph_plots=graph_plots,
overview_plots=overview_plots, legend=legend,
baseline_plots=baseline_plots)
def run_test(self, opts):
# Resolve the cc_under_test path.
opts.cc = resolve_command_path(opts.cc)
if not lnt.testing.util.compilers.is_valid(opts.cc):
self._fatal('--cc does not point to a valid executable.')
# Attempt to infer the cxx compiler if not given.
if opts.cc and opts.cxx is None:
opts.cxx = lnt.testing.util.compilers.infer_cxx_compiler(opts.cc)
if opts.cxx is not None:
logger.info("inferred C++ compiler under test as: %r" %
(opts.cxx, ))
if opts.cxx is None:
self._fatal('--cxx is required (and could not be inferred)')
# Force the CC and CXX variables to be absolute paths.
cc_abs = os.path.abspath(commands.which(opts.cc))
cxx_abs = os.path.abspath(commands.which(opts.cxx))
if not os.path.exists(cc_abs):
self._fatal("unable to determine absolute path for --cc: %r" %
(opts.cc, ))
if not os.path.exists(cxx_abs):
self._fatal("unable to determine absolute path for --cc: %r" %
(opts.cc, ))
opts.cc = cc_abs
opts.cxx = cxx_abs
# If no ld was set, set ld to opts.cc
if opts.ld is None:
opts.ld = opts.cc
# If no ldxx was set, set ldxx to opts.cxx
if opts.ldxx is None:
opts.ldxx = opts.cxx
# Set up the sandbox.
global g_output_dir
if not os.path.exists(opts.sandbox_path):
print >> sys.stderr, "%s: creating sandbox: %r" % (
timestamp(), opts.sandbox_path)
os.mkdir(opts.sandbox_path)
if opts.timestamp_build:
fmt_timestamp = timestamp().replace(' ', '_').replace(':', '-')
report_name = "test-%s" % (fmt_timestamp)
else:
report_name = "build"
g_output_dir = os.path.join(os.path.abspath(opts.sandbox_path),
report_name)
try:
os.mkdir(g_output_dir)
except OSError as e:
if e.errno == errno.EEXIST:
self._fatal("sandbox output directory %r already exists!" %
(g_output_dir, ))
else:
raise
# Setup log file
global g_log
def setup_log(output_dir):
def stderr_log_handler():
h = logging.StreamHandler()
f = logging.Formatter(
"%(asctime)-7s: %(levelname)s: %(message)s",
"%Y-%m-%d %H:%M:%S")
h.setFormatter(f)
return h
def file_log_handler(path):
h = logging.FileHandler(path, mode='w')
f = logging.Formatter(
"%(asctime)-7s: %(levelname)s: %(message)s",
"%Y-%m-%d %H:%M:%S")
h.setFormatter(f)
return h
log = logging.Logger('compile_test')
log.setLevel(logging.INFO)
log.addHandler(
file_log_handler(os.path.join(output_dir, 'test.log')))
log.addHandler(stderr_log_handler())
return log
g_log = setup_log(g_output_dir)
# Collect machine and run information.
machine_info, run_info = machineinfo.get_machine_information(
opts.use_machdep_info)
# FIXME: Include information on test source versions.
#
# FIXME: Get more machine information? Cocoa.h hash, for example.
for name, cmd in (('sys_cc_version', ('/usr/bin/gcc', '-v')),
('sys_as_version', ('/usr/bin/as', '-v',
'/dev/null')),
('sys_ld_version', ('/usr/bin/ld', '-v')),
('sys_xcodebuild', ('xcodebuild', '-version'))):
run_info[name] = commands.capture(cmd, include_stderr=True).strip()
# Set command line machine and run information.
for info, params in ((machine_info, opts.machine_parameters),
(run_info, opts.run_parameters)):
for entry in params:
if '=' not in entry:
name, value = entry, ''
else:
name, value = entry.split('=', 1)
info[name] = value
# Set user variables.
variables = {}
variables['cc'] = opts.cc
variables['run_count'] = opts.run_count
# Get compiler info.
cc_info = lnt.testing.util.compilers.get_cc_info(variables['cc'])
variables.update(cc_info)
# Set the run order from the user, if given.
if opts.run_order is not None:
variables['run_order'] = opts.run_order
else:
# Otherwise, use the inferred run order.
variables['run_order'] = cc_info['inferred_run_order']
logger.info("inferred run order to be: %r" %
(variables['run_order'], ))
if opts.verbose:
format = pprint.pformat(variables)
msg = '\n\t'.join(['using variables:'] + format.splitlines())
logger.info(msg)
format = pprint.pformat(machine_info)
msg = '\n\t'.join(['using machine info:'] + format.splitlines())
logger.info(msg)
format = pprint.pformat(run_info)
msg = '\n\t'.join(['using run info:'] + format.splitlines())
logger.info(msg)
# Compute the set of flags to test.
if not opts.flags_to_test:
flags_to_test = DEFAULT_FLAGS_TO_TEST
else:
flags_to_test = [
string.split(' ') for string in opts.flags_to_test
]
# Compute the set of job counts to use in full build tests.
if not opts.jobs_to_test:
jobs_to_test = [1, 2, 4, 8]
else:
jobs_to_test = opts.jobs_to_test
# Compute the build configurations to test.
if not opts.configs_to_test:
configs_to_test = ['Debug', 'Release']
else:
configs_to_test = opts.configs_to_test
# Compute the list of all tests.
all_tests = list(
get_tests(opts.test_suite_externals, opts.test_subdir,
flags_to_test, jobs_to_test, configs_to_test))
# Show the tests, if requested.
if opts.show_tests:
print >> sys.stderr, 'Available Tests'
for name in sorted(set(name for name, _ in all_tests)):
print >> sys.stderr, ' %s' % (name, )
print
raise SystemExit
# Find the tests to run.
if not opts.tests and not opts.test_filters:
tests_to_run = list(all_tests)
else:
all_test_names = set(test[0] for test in all_tests)
# Validate the test names.
requested_tests = set(opts.tests)
missing_tests = requested_tests - all_test_names
if missing_tests:
self._fatal(("invalid test names %s, use --show-tests to "
"see available tests") %
(", ".join(map(repr, missing_tests)), ))
# Validate the test filters.
test_filters = [
re.compile(pattern) for pattern in opts.test_filters
]
# Form the list of tests.
tests_to_run = [
test for test in all_tests if
(test[0] in requested_tests or
[True for filter in test_filters if filter.search(test[0])])
]
if not tests_to_run:
self._fatal("no tests requested "
"(invalid --test or --test-filter options)!")
# Ensure output directory is available.
if not os.path.exists(g_output_dir):
os.mkdir(g_output_dir)
# Execute the run.
run_info.update(variables)
run_info['tag'] = tag = 'compile'
testsamples = []
start_time = datetime.utcnow()
g_log.info('run started')
g_log.info('using CC: %r' % opts.cc)
g_log.info('using CXX: %r' % opts.cxx)
no_errors = True
for basename, test_fn in tests_to_run:
for success, name, samples in test_fn(basename, run_info,
variables):
g_log.info('collected samples: %r' % name)
num_samples = len(samples)
if num_samples:
samples_median = '%.4f' % (stats.median(samples), )
samples_mad = '%.4f' % (
stats.median_absolute_deviation(samples), )
else:
samples_median = samples_mad = 'N/A'
g_log.info('N=%d, median=%s, MAD=%s' %
(num_samples, samples_median, samples_mad))
test_name = '%s.%s' % (tag, name)
if not success:
testsamples.append(
lnt.testing.TestSamples(test_name + '.status',
[lnt.testing.FAIL]))
no_errors = False
if samples:
testsamples.append(
lnt.testing.TestSamples(test_name, samples))
run_info['no_errors'] = no_errors
end_time = datetime.utcnow()
g_log.info('run complete')
# Package up the report.
machine = lnt.testing.Machine(opts.machine_name, machine_info)
run = lnt.testing.Run(start_time, end_time, info=run_info)
# Write out the report.
lnt_report_path = os.path.join(g_output_dir, 'report.json')
report = lnt.testing.Report(machine, run, testsamples)
# Save report to disk for submission.
self.print_report(report, lnt_report_path)
# Then, also print to screen if requested.
if opts.output is not None:
self.print_report(report, opts.output)
server_report = self.submit(lnt_report_path, opts, ts_name='compile')
return server_report
class CompileTest(builtintest.BuiltinTest):
def describe(self):
return 'Single file compile-time performance testing'
def run_test(self, name, args):
global opts
parser = OptionParser(
("%(name)s [options] [<output file>]\n" + usage_info) % locals())
parser.add_option("-s",
"--sandbox",
dest="sandbox_path",
help="Parent directory to build and run tests in",
type=str,
default=None,
metavar="PATH")
group = OptionGroup(parser, "Test Options")
group.add_option("",
"--no-timestamp",
dest="timestamp_build",
help="Don't timestamp build directory (for testing)",
action="store_false",
default=True)
group.add_option("",
"--cc",
dest="cc",
type='str',
help="Path to the compiler under test",
action="store",
default=None)
group.add_option("",
"--cxx",
dest="cxx",
help="Path to the C++ compiler to test",
type=str,
default=None)
group.add_option(
"",
"--ld",
dest="ld",
help="Path to the c linker to use. (Xcode Distinction)",
type=str,
default=None)
group.add_option(
"",
"--ldxx",
dest="ldxx",
help="Path to the cxx linker to use. (Xcode Distinction)",
type=str,
default=None)
group.add_option("",
"--test-externals",
dest="test_suite_externals",
help="Path to the LLVM test-suite externals",
type=str,
default=None,
metavar="PATH")
group.add_option("",
"--machine-param",
dest="machine_parameters",
metavar="NAME=VAL",
help="Add 'NAME' = 'VAL' to the machine parameters",
type=str,
action="append",
default=[])
group.add_option("",
"--run-param",
dest="run_parameters",
metavar="NAME=VAL",
help="Add 'NAME' = 'VAL' to the run parameters",
type=str,
action="append",
default=[])
group.add_option("",
"--run-order",
dest="run_order",
metavar="STR",
help="String to use to identify and order this run",
action="store",
type=str,
default=None)
group.add_option(
"",
"--test-subdir",
dest="test_subdir",
help="Subdirectory of test external dir to look for tests in.",
type=str,
default="lnt-compile-suite-src")
parser.add_option_group(group)
group = OptionGroup(parser, "Test Selection")
group.add_option("",
"--no-memory-profiling",
dest="memory_profiling",
help="Disable memory profiling",
action="store_false",
default=True)
group.add_option("",
"--multisample",
dest="run_count",
metavar="N",
help="Accumulate test data from multiple runs",
action="store",
type=int,
default=3)
group.add_option("",
"--min-sample-time",
dest="min_sample_time",
help="Ensure all tests run for at least N seconds",
metavar="N",
action="store",
type=float,
default=.5)
group.add_option("",
"--save-temps",
dest="save_temps",
help="Save temporary build output files",
action="store_true",
default=False)
group.add_option(
"",
"--show-tests",
dest="show_tests",
help="Only list the availables tests that will be run",
action="store_true",
default=False)
group.add_option("",
"--test",
dest="tests",
metavar="NAME",
help="Individual test to run",
action="append",
default=[])
group.add_option("",
"--test-filter",
dest="test_filters",
help="Run tests matching the given pattern",
metavar="REGEXP",
action="append",
default=[])
group.add_option("",
"--flags-to-test",
dest="flags_to_test",
help="Add a set of flags to test (space separated)",
metavar="FLAGLIST",
action="append",
default=[])
group.add_option("",
"--jobs-to-test",
dest="jobs_to_test",
help="Add a job count to test (full builds)",
metavar="NUM",
action="append",
default=[],
type=int)
group.add_option("",
"--config-to-test",
dest="configs_to_test",
help="Add build configuration to test (full builds)",
metavar="NAME",
action="append",
default=[],
choices=('Debug', 'Release'))
parser.add_option_group(group)
group = OptionGroup(parser, "Output Options")
group.add_option("",
"--no-machdep-info",
dest="use_machdep_info",
help=("Don't put machine (instance) dependent "
"variables in machine info"),
action="store_false",
default=True)
group.add_option("",
"--machine-name",
dest="machine_name",
type='str',
help="Machine name to use in submission [%default]",
action="store",
default=platform.uname()[1])
group.add_option(
"",
"--submit",
dest="submit_url",
metavar="URLORPATH",
help=("autosubmit the test result to the given server "
"(or local instance) [%default]"),
type=str,
default=None)
group.add_option(
"",
"--commit",
dest="commit",
help=("whether the autosubmit result should be committed "
"[%default]"),
type=int,
default=True)
group.add_option(
"",
"--output",
dest="output",
metavar="PATH",
help="write raw report data to PATH (or stdout if '-')",
action="store",
default=None)
group.add_option("-v",
"--verbose",
dest="verbose",
help="show verbose test results",
action="store_true",
default=False)
parser.add_option_group(group)
opts, args = parser.parse_args(args)
if len(args) != 0:
parser.error("invalid number of arguments")
if opts.cc is None:
parser.error("You must specify a --cc argument.")
# Resolve the cc_under_test path.
opts.cc = resolve_command_path(opts.cc)
if not lnt.testing.util.compilers.is_valid(opts.cc):
parser.error('--cc does not point to a valid executable.')
# Attempt to infer the cxx compiler if not given.
if opts.cc and opts.cxx is None:
opts.cxx = lnt.testing.util.compilers.infer_cxx_compiler(opts.cc)
if opts.cxx is not None:
note("inferred C++ compiler under test as: %r" % (opts.cxx, ))
# Validate options.
if opts.cc is None:
parser.error('--cc is required')
if opts.cxx is None:
parser.error('--cxx is required (and could not be inferred)')
if opts.sandbox_path is None:
parser.error('--sandbox is required')
if opts.test_suite_externals is None:
parser.error("--test-externals option is required")
# Force the CC and CXX variables to be absolute paths.
cc_abs = os.path.abspath(commands.which(opts.cc))
cxx_abs = os.path.abspath(commands.which(opts.cxx))
if not os.path.exists(cc_abs):
parser.error("unable to determine absolute path for --cc: %r" %
(opts.cc, ))
if not os.path.exists(cxx_abs):
parser.error("unable to determine absolute path for --cc: %r" %
(opts.cc, ))
opts.cc = cc_abs
opts.cxx = cxx_abs
# If no ld was set, set ld to opts.cc
if opts.ld is None:
opts.ld = opts.cc
# If no ldxx was set, set ldxx to opts.cxx
if opts.ldxx is None:
opts.ldxx = opts.cxx
# Set up the sandbox.
global g_output_dir
if not os.path.exists(opts.sandbox_path):
print >> sys.stderr, "%s: creating sandbox: %r" % (
timestamp(), opts.sandbox_path)
os.mkdir(opts.sandbox_path)
if opts.timestamp_build:
report_name = "test-%s" % (timestamp().replace(' ', '_').replace(
':', '-'))
else:
report_name = "build"
g_output_dir = os.path.join(os.path.abspath(opts.sandbox_path),
report_name)
try:
os.mkdir(g_output_dir)
except OSError, e:
if e.errno == errno.EEXIST:
parser.error("sandbox output directory %r already exists!" %
(g_output_dir, ))
else:
raise
# Setup log file
global g_log
def setup_log(output_dir):
def stderr_log_handler():
h = logging.StreamHandler()
f = logging.Formatter(
"%(asctime)-7s: %(levelname)s: %(message)s",
"%Y-%m-%d %H:%M:%S")
h.setFormatter(f)
return h
def file_log_handler(path):
h = logging.FileHandler(path, mode='w')
f = logging.Formatter(
"%(asctime)-7s: %(levelname)s: %(message)s",
"%Y-%m-%d %H:%M:%S")
h.setFormatter(f)
return h
l = logging.Logger('compile_test')
l.setLevel(logging.INFO)
l.addHandler(file_log_handler(os.path.join(output_dir,
'test.log')))
l.addHandler(stderr_log_handler())
return l
g_log = setup_log(g_output_dir)
# Collect machine and run information.
machine_info, run_info = machineinfo.get_machine_information(
opts.use_machdep_info)
# FIXME: Include information on test source versions.
#
# FIXME: Get more machine information? Cocoa.h hash, for example.
for name, cmd in (('sys_cc_version', ('/usr/bin/gcc', '-v')),
('sys_as_version', ('/usr/bin/as', '-v',
'/dev/null')),
('sys_ld_version', ('/usr/bin/ld', '-v')),
('sys_xcodebuild', ('xcodebuild', '-version'))):
run_info[name] = commands.capture(cmd, include_stderr=True).strip()
# Set command line machine and run information.
for info, params in ((machine_info, opts.machine_parameters),
(run_info, opts.run_parameters)):
for entry in params:
if '=' not in entry:
name, value = entry, ''
else:
name, value = entry.split('=', 1)
info[name] = value
# Set user variables.
variables = {}
variables['cc'] = opts.cc
variables['run_count'] = opts.run_count
# Get compiler info.
cc_info = lnt.testing.util.compilers.get_cc_info(variables['cc'])
variables.update(cc_info)
# Set the run order from the user, if given.
if opts.run_order is not None:
variables['run_order'] = opts.run_order
else:
# Otherwise, use the inferred run order.
variables['run_order'] = cc_info['inferred_run_order']
note("inferred run order to be: %r" % (variables['run_order'], ))
if opts.verbose:
format = pprint.pformat(variables)
msg = '\n\t'.join(['using variables:'] + format.splitlines())
note(msg)
format = pprint.pformat(machine_info)
msg = '\n\t'.join(['using machine info:'] + format.splitlines())
note(msg)
format = pprint.pformat(run_info)
msg = '\n\t'.join(['using run info:'] + format.splitlines())
note(msg)
# Compute the set of flags to test.
if not opts.flags_to_test:
flags_to_test = [('-O0', ), (
'-O0',
'-g',
), ('-Os', '-g'), ('-O3', )]
else:
flags_to_test = [
string.split(' ') for string in opts.flags_to_test
]
# Compute the set of job counts to use in full build tests.
if not opts.jobs_to_test:
jobs_to_test = [1, 2, 4, 8]
else:
jobs_to_test = opts.jobs_to_test
# Compute the build configurations to test.
if not opts.configs_to_test:
configs_to_test = ['Debug', 'Release']
else:
configs_to_test = opts.configs_to_test
# Compute the list of all tests.
all_tests = list(
get_tests(opts.test_suite_externals, opts.test_subdir,
flags_to_test, jobs_to_test, configs_to_test))
# Show the tests, if requested.
if opts.show_tests:
print >> sys.stderr, 'Available Tests'
for name in sorted(set(name for name, _ in all_tests)):
print >> sys.stderr, ' %s' % (name, )
print
raise SystemExit
# Find the tests to run.
if not opts.tests and not opts.test_filters:
tests_to_run = list(all_tests)
else:
all_test_names = set(test[0] for test in all_tests)
# Validate the test names.
requested_tests = set(opts.tests)
missing_tests = requested_tests - all_test_names
if missing_tests:
parser.error(("invalid test names %s, use --show-tests to "
"see available tests") %
(", ".join(map(repr, missing_tests)), ))
# Validate the test filters.
test_filters = [
re.compile(pattern) for pattern in opts.test_filters
]
# Form the list of tests.
tests_to_run = [
test for test in all_tests if
(test[0] in requested_tests or
[True for filter in test_filters if filter.search(test[0])])
]
if not tests_to_run:
parser.error(
"no tests requested (invalid --test or --test-filter options)!"
)
# Ensure output directory is available.
if not os.path.exists(g_output_dir):
os.mkdir(g_output_dir)
# Execute the run.
run_info.update(variables)
run_info['tag'] = tag = 'compile'
testsamples = []
start_time = datetime.utcnow()
g_log.info('run started')
g_log.info('using CC: %r' % opts.cc)
g_log.info('using CXX: %r' % opts.cxx)
for basename, test_fn in tests_to_run:
for success, name, samples in test_fn(basename, run_info,
variables):
g_log.info('collected samples: %r' % name)
num_samples = len(samples)
if num_samples:
samples_median = '%.4f' % (stats.median(samples), )
samples_mad = '%.4f' % (
stats.median_absolute_deviation(samples), )
else:
samples_median = samples_mad = 'N/A'
g_log.info('N=%d, median=%s, MAD=%s' %
(num_samples, samples_median, samples_mad))
test_name = '%s.%s' % (tag, name)
if not success:
testsamples.append(
lnt.testing.TestSamples(test_name + '.status',
[lnt.testing.FAIL]))
if samples:
testsamples.append(
lnt.testing.TestSamples(test_name, samples))
end_time = datetime.utcnow()
g_log.info('run complete')
# Package up the report.
machine = lnt.testing.Machine(opts.machine_name, machine_info)
run = lnt.testing.Run(start_time, end_time, info=run_info)
# Write out the report.
lnt_report_path = os.path.join(g_output_dir, 'report.json')
report = lnt.testing.Report(machine, run, testsamples)
# Save report to disk for submission.
self.print_report(report, lnt_report_path)
# Then, also print to screen if requested.
if opts.output is not None:
self.print_report(report, opts.output)
server_report = self.submit(lnt_report_path, opts)
return server_report
def __init__(self,
aggregation_fn,
cur_failed,
prev_failed,
samples,
prev_samples,
cur_hash,
prev_hash,
cur_profile=None,
prev_profile=None,
confidence_lv=0.05,
bigger_is_better=False,
stable_test=True):
self.aggregation_fn = aggregation_fn
# Special case: if we're using the minimum to aggregate, swap it for max
# if bigger_is_better.
if aggregation_fn == stats.safe_min and bigger_is_better:
aggregation_fn = stats.safe_max
self.cur_hash = cur_hash
self.prev_hash = prev_hash
self.cur_profile = cur_profile
self.prev_profile = prev_profile
if samples:
self.current = aggregation_fn(samples)
else:
self.current = None
self.previous = None
# Compute the comparison status for the test value.
self.delta = 0
self.pct_delta = 0.0
if self.current and prev_samples:
prev_median = stats.median(prev_samples)
self.delta, value = absmin_diff(self.current, [prev_median])
if value != 0:
self.pct_delta = self.delta / value
self.previous = value
# If we have multiple values for this run, use that to estimate the
# distribution.
#
# We can get integer sample types here - for example if the field is
# .exec.status. Make sure we don't assert by avoiding the stats
# functions in this case.
if samples and len(samples) > 1 and isinstance(samples[0], float):
self.stddev = stats.standard_deviation(samples)
self.MAD = stats.median_absolute_deviation(samples)
self.variance = stats.variance(samples)
else:
self.stddev = None
self.MAD = None
self.variance = None
if prev_samples and len(prev_samples) > 1 and isinstance(
prev_samples[0], float):
self.prev_stddev = stats.standard_deviation(prev_samples)
self.prev_MAD = stats.median_absolute_deviation(prev_samples)
self.prev_variance = stats.variance(prev_samples)
else:
self.prev_stddev = None
self.prev_MAD = None
self.prev_variance = None
self.stddev_mean = None # Only calculate this if needed.
self.failed = cur_failed
self.prev_failed = prev_failed
self.samples = samples
self.prev_samples = prev_samples
self.confidence_lv = confidence_lv
self.bigger_is_better = bigger_is_better
self.stable_test = stable_test
def run_test(self, opts):
# Resolve the cc_under_test path.
opts.cc = resolve_command_path(opts.cc)
if not lnt.testing.util.compilers.is_valid(opts.cc):
self._fatal('--cc does not point to a valid executable.')
# Attempt to infer the cxx compiler if not given.
if opts.cc and opts.cxx is None:
opts.cxx = lnt.testing.util.compilers.infer_cxx_compiler(opts.cc)
if opts.cxx is not None:
logger.info("inferred C++ compiler under test as: %r" %
(opts.cxx,))
if opts.cxx is None:
self._fatal('--cxx is required (and could not be inferred)')
# Force the CC and CXX variables to be absolute paths.
cc_abs = os.path.abspath(commands.which(opts.cc))
cxx_abs = os.path.abspath(commands.which(opts.cxx))
if not os.path.exists(cc_abs):
self._fatal("unable to determine absolute path for --cc: %r" % (
opts.cc,))
if not os.path.exists(cxx_abs):
self._fatal("unable to determine absolute path for --cc: %r" % (
opts.cc,))
opts.cc = cc_abs
opts.cxx = cxx_abs
# If no ld was set, set ld to opts.cc
if opts.ld is None:
opts.ld = opts.cc
# If no ldxx was set, set ldxx to opts.cxx
if opts.ldxx is None:
opts.ldxx = opts.cxx
# Set up the sandbox.
global g_output_dir
if not os.path.exists(opts.sandbox_path):
print >>sys.stderr, "%s: creating sandbox: %r" % (
timestamp(), opts.sandbox_path)
os.mkdir(opts.sandbox_path)
if opts.timestamp_build:
fmt_timestamp = timestamp().replace(' ', '_').replace(':', '-')
report_name = "test-%s" % (fmt_timestamp)
else:
report_name = "build"
g_output_dir = os.path.join(os.path.abspath(opts.sandbox_path),
report_name)
try:
os.mkdir(g_output_dir)
except OSError as e:
if e.errno == errno.EEXIST:
self._fatal("sandbox output directory %r already exists!" % (
g_output_dir,))
else:
raise
# Setup log file
global g_log
def setup_log(output_dir):
def stderr_log_handler():
h = logging.StreamHandler()
f = logging.Formatter(
"%(asctime)-7s: %(levelname)s: %(message)s",
"%Y-%m-%d %H:%M:%S")
h.setFormatter(f)
return h
def file_log_handler(path):
h = logging.FileHandler(path, mode='w')
f = logging.Formatter(
"%(asctime)-7s: %(levelname)s: %(message)s",
"%Y-%m-%d %H:%M:%S")
h.setFormatter(f)
return h
log = logging.Logger('compile_test')
log.setLevel(logging.INFO)
log.addHandler(file_log_handler(os.path.join(output_dir,
'test.log')))
log.addHandler(stderr_log_handler())
return log
g_log = setup_log(g_output_dir)
# Collect machine and run information.
machine_info, run_info = machineinfo.get_machine_information(
opts.use_machdep_info)
# FIXME: Include information on test source versions.
#
# FIXME: Get more machine information? Cocoa.h hash, for example.
for name, cmd in (('sys_cc_version', ('/usr/bin/gcc', '-v')),
('sys_as_version',
('/usr/bin/as', '-v', '/dev/null')),
('sys_ld_version', ('/usr/bin/ld', '-v')),
('sys_xcodebuild', ('xcodebuild', '-version'))):
run_info[name] = commands.capture(cmd, include_stderr=True).strip()
# Set command line machine and run information.
for info, params in ((machine_info, opts.machine_parameters),
(run_info, opts.run_parameters)):
for entry in params:
if '=' not in entry:
name, value = entry, ''
else:
name, value = entry.split('=', 1)
info[name] = value
# Set user variables.
variables = {}
variables['cc'] = opts.cc
variables['run_count'] = opts.run_count
# Get compiler info.
cc_info = lnt.testing.util.compilers.get_cc_info(variables['cc'])
variables.update(cc_info)
# Set the run order from the user, if given.
if opts.run_order is not None:
variables['run_order'] = opts.run_order
else:
# Otherwise, use the inferred run order.
variables['run_order'] = cc_info['inferred_run_order']
logger.info("inferred run order to be: %r" %
(variables['run_order'],))
if opts.verbose:
format = pprint.pformat(variables)
msg = '\n\t'.join(['using variables:'] + format.splitlines())
logger.info(msg)
format = pprint.pformat(machine_info)
msg = '\n\t'.join(['using machine info:'] + format.splitlines())
logger.info(msg)
format = pprint.pformat(run_info)
msg = '\n\t'.join(['using run info:'] + format.splitlines())
logger.info(msg)
# Compute the set of flags to test.
if not opts.flags_to_test:
flags_to_test = DEFAULT_FLAGS_TO_TEST
else:
flags_to_test = [string.split(' ')
for string in opts.flags_to_test]
# Compute the set of job counts to use in full build tests.
if not opts.jobs_to_test:
jobs_to_test = [1, 2, 4, 8]
else:
jobs_to_test = opts.jobs_to_test
# Compute the build configurations to test.
if not opts.configs_to_test:
configs_to_test = ['Debug', 'Release']
else:
configs_to_test = opts.configs_to_test
# Compute the list of all tests.
all_tests = list(get_tests(opts.test_suite_externals, opts.test_subdir,
flags_to_test, jobs_to_test,
configs_to_test))
# Show the tests, if requested.
if opts.show_tests:
print >>sys.stderr, 'Available Tests'
for name in sorted(set(name for name, _ in all_tests)):
print >>sys.stderr, ' %s' % (name,)
print
raise SystemExit
# Find the tests to run.
if not opts.tests and not opts.test_filters:
tests_to_run = list(all_tests)
else:
all_test_names = set(test[0] for test in all_tests)
# Validate the test names.
requested_tests = set(opts.tests)
missing_tests = requested_tests - all_test_names
if missing_tests:
self._fatal(("invalid test names %s, use --show-tests to "
"see available tests") %
(", ".join(map(repr, missing_tests)), ))
# Validate the test filters.
test_filters = [re.compile(pattern)
for pattern in opts.test_filters]
# Form the list of tests.
tests_to_run = [test
for test in all_tests
if (test[0] in requested_tests or
[True
for filter in test_filters
if filter.search(test[0])])]
if not tests_to_run:
self._fatal("no tests requested "
"(invalid --test or --test-filter options)!")
# Ensure output directory is available.
if not os.path.exists(g_output_dir):
os.mkdir(g_output_dir)
# Execute the run.
run_info.update(variables)
run_info['tag'] = tag = 'compile'
testsamples = []
start_time = datetime.utcnow()
g_log.info('run started')
g_log.info('using CC: %r' % opts.cc)
g_log.info('using CXX: %r' % opts.cxx)
no_errors = True
for basename, test_fn in tests_to_run:
for success, name, samples in test_fn(basename, run_info,
variables):
g_log.info('collected samples: %r' % name)
num_samples = len(samples)
if num_samples:
samples_median = '%.4f' % (stats.median(samples),)
samples_mad = '%.4f' % (
stats.median_absolute_deviation(samples),)
else:
samples_median = samples_mad = 'N/A'
g_log.info('N=%d, median=%s, MAD=%s' % (
num_samples, samples_median, samples_mad))
test_name = '%s.%s' % (tag, name)
if not success:
testsamples.append(lnt.testing.TestSamples(
test_name + '.status', [lnt.testing.FAIL]))
no_errors = False
if samples:
testsamples.append(lnt.testing.TestSamples(
test_name, samples))
run_info['no_errors'] = no_errors
end_time = datetime.utcnow()
g_log.info('run complete')
# Package up the report.
machine = lnt.testing.Machine(opts.machine_name, machine_info)
run = lnt.testing.Run(start_time, end_time, info=run_info)
# Write out the report.
lnt_report_path = os.path.join(g_output_dir, 'report.json')
report = lnt.testing.Report(machine, run, testsamples)
# Save report to disk for submission.
self.print_report(report, lnt_report_path)
# Then, also print to screen if requested.
if opts.output is not None:
self.print_report(report, opts.output)
server_report = self.submit(lnt_report_path, opts, ts_name='compile')
return server_report
def run_test(self, name, args):
global opts
parser = OptionParser(
("%(name)s [options] [<output file>]\n" +
usage_info) % locals())
parser.add_option("-s", "--sandbox", dest="sandbox_path",
help="Parent directory to build and run tests in",
type=str, default=None, metavar="PATH")
group = OptionGroup(parser, "Test Options")
group.add_option("", "--no-timestamp", dest="timestamp_build",
help="Don't timestamp build directory (for testing)",
action="store_false", default=True)
group.add_option("", "--cc", dest="cc", type='str',
help="Path to the compiler under test",
action="store", default=None)
group.add_option("", "--cxx", dest="cxx",
help="Path to the C++ compiler to test",
type=str, default=None)
group.add_option("", "--ld", dest="ld",
help="Path to the c linker to use. (Xcode Distinction)",
type=str, default=None)
group.add_option("", "--ldxx", dest="ldxx",
help="Path to the cxx linker to use. (Xcode Distinction)",
type=str, default=None)
group.add_option("", "--test-externals", dest="test_suite_externals",
help="Path to the LLVM test-suite externals",
type=str, default=None, metavar="PATH")
group.add_option("", "--machine-param", dest="machine_parameters",
metavar="NAME=VAL",
help="Add 'NAME' = 'VAL' to the machine parameters",
type=str, action="append", default=[])
group.add_option("", "--run-param", dest="run_parameters",
metavar="NAME=VAL",
help="Add 'NAME' = 'VAL' to the run parameters",
type=str, action="append", default=[])
group.add_option("", "--run-order", dest="run_order", metavar="STR",
help="String to use to identify and order this run",
action="store", type=str, default=None)
group.add_option("", "--test-subdir", dest="test_subdir",
help="Subdirectory of test external dir to look for tests in.",
type=str, default="lnt-compile-suite-src")
parser.add_option_group(group)
group = OptionGroup(parser, "Test Selection")
group.add_option("", "--no-memory-profiling", dest="memory_profiling",
help="Disable memory profiling",
action="store_false", default=True)
group.add_option("", "--multisample", dest="run_count", metavar="N",
help="Accumulate test data from multiple runs",
action="store", type=int, default=3)
group.add_option("", "--min-sample-time", dest="min_sample_time",
help="Ensure all tests run for at least N seconds",
metavar="N", action="store", type=float, default=.5)
group.add_option("", "--save-temps", dest="save_temps",
help="Save temporary build output files",
action="store_true", default=False)
group.add_option("", "--show-tests", dest="show_tests",
help="Only list the availables tests that will be run",
action="store_true", default=False)
group.add_option("", "--test", dest="tests", metavar="NAME",
help="Individual test to run",
action="append", default=[])
group.add_option("", "--test-filter", dest="test_filters",
help="Run tests matching the given pattern",
metavar="REGEXP", action="append", default=[])
group.add_option("", "--flags-to-test", dest="flags_to_test",
help="Add a set of flags to test (space separated)",
metavar="FLAGLIST", action="append", default=[])
group.add_option("", "--jobs-to-test", dest="jobs_to_test",
help="Add a job count to test (full builds)",
metavar="NUM", action="append", default=[], type=int)
group.add_option("", "--config-to-test", dest="configs_to_test",
help="Add build configuration to test (full builds)",
metavar="NAME", action="append", default=[],
choices=('Debug', 'Release'))
parser.add_option_group(group)
group = OptionGroup(parser, "Output Options")
group.add_option("", "--no-machdep-info", dest="use_machdep_info",
help=("Don't put machine (instance) dependent "
"variables in machine info"),
action="store_false", default=True)
group.add_option("", "--machine-name", dest="machine_name", type='str',
help="Machine name to use in submission [%default]",
action="store", default=platform.uname()[1])
group.add_option("", "--submit", dest="submit_url", metavar="URLORPATH",
help=("autosubmit the test result to the given server "
"(or local instance) [%default]"),
type=str, default=None)
group.add_option("", "--commit", dest="commit",
help=("whether the autosubmit result should be committed "
"[%default]"),
type=int, default=True)
group.add_option("", "--output", dest="output", metavar="PATH",
help="write raw report data to PATH (or stdout if '-')",
action="store", default=None)
group.add_option("-v", "--verbose", dest="verbose",
help="show verbose test results",
action="store_true", default=False)
parser.add_option_group(group)
opts, args = parser.parse_args(args)
if len(args) != 0:
parser.error("invalid number of arguments")
if opts.cc is None:
parser.error("You must specify a --cc argument.")
# Resolve the cc_under_test path.
opts.cc = resolve_command_path(opts.cc)
if not lnt.testing.util.compilers.is_valid(opts.cc):
parser.error('--cc does not point to a valid executable.')
# Attempt to infer the cxx compiler if not given.
if opts.cc and opts.cxx is None:
opts.cxx = lnt.testing.util.compilers.infer_cxx_compiler(opts.cc)
if opts.cxx is not None:
note("inferred C++ compiler under test as: %r" % (opts.cxx,))
# Validate options.
if opts.cc is None:
parser.error('--cc is required')
if opts.cxx is None:
parser.error('--cxx is required (and could not be inferred)')
if opts.sandbox_path is None:
parser.error('--sandbox is required')
if opts.test_suite_externals is None:
parser.error("--test-externals option is required")
# Force the CC and CXX variables to be absolute paths.
cc_abs = os.path.abspath(commands.which(opts.cc))
cxx_abs = os.path.abspath(commands.which(opts.cxx))
if not os.path.exists(cc_abs):
parser.error("unable to determine absolute path for --cc: %r" % (
opts.cc,))
if not os.path.exists(cxx_abs):
parser.error("unable to determine absolute path for --cc: %r" % (
opts.cc,))
opts.cc = cc_abs
opts.cxx = cxx_abs
# If no ld was set, set ld to opts.cc
if opts.ld is None:
opts.ld = opts.cc
# If no ldxx was set, set ldxx to opts.cxx
if opts.ldxx is None:
opts.ldxx = opts.cxx
# Set up the sandbox.
global g_output_dir
if not os.path.exists(opts.sandbox_path):
print >>sys.stderr, "%s: creating sandbox: %r" % (
timestamp(), opts.sandbox_path)
os.mkdir(opts.sandbox_path)
if opts.timestamp_build:
fmt_timestamp = timestamp().replace(' ', '_').replace(':', '-')
report_name = "test-%s" % (fmt_timestamp)
else:
report_name = "build"
g_output_dir = os.path.join(os.path.abspath(opts.sandbox_path),
report_name)
try:
os.mkdir(g_output_dir)
except OSError(e):
if e.errno == errno.EEXIST:
parser.error("sandbox output directory %r already exists!" % (
g_output_dir,))
else:
raise
# Setup log file
global g_log
def setup_log(output_dir):
def stderr_log_handler():
h = logging.StreamHandler()
f = logging.Formatter("%(asctime)-7s: %(levelname)s: %(message)s",
"%Y-%m-%d %H:%M:%S")
h.setFormatter(f)
return h
def file_log_handler(path):
h = logging.FileHandler(path, mode='w')
f = logging.Formatter("%(asctime)-7s: %(levelname)s: %(message)s",
"%Y-%m-%d %H:%M:%S")
h.setFormatter(f)
return h
l = logging.Logger('compile_test')
l.setLevel(logging.INFO)
l.addHandler(file_log_handler(os.path.join(output_dir, 'test.log')))
l.addHandler(stderr_log_handler())
return l
g_log = setup_log(g_output_dir)
# Collect machine and run information.
machine_info, run_info = machineinfo.get_machine_information(
opts.use_machdep_info)
# FIXME: Include information on test source versions.
#
# FIXME: Get more machine information? Cocoa.h hash, for example.
for name, cmd in (('sys_cc_version', ('/usr/bin/gcc', '-v')),
('sys_as_version', ('/usr/bin/as', '-v', '/dev/null')),
('sys_ld_version', ('/usr/bin/ld', '-v')),
('sys_xcodebuild', ('xcodebuild', '-version'))):
run_info[name] = commands.capture(cmd, include_stderr=True).strip()
# Set command line machine and run information.
for info, params in ((machine_info, opts.machine_parameters),
(run_info, opts.run_parameters)):
for entry in params:
if '=' not in entry:
name, value = entry, ''
else:
name, value = entry.split('=', 1)
info[name] = value
# Set user variables.
variables = {}
variables['cc'] = opts.cc
variables['run_count'] = opts.run_count
# Get compiler info.
cc_info = lnt.testing.util.compilers.get_cc_info(variables['cc'])
variables.update(cc_info)
# Set the run order from the user, if given.
if opts.run_order is not None:
variables['run_order'] = opts.run_order
else:
# Otherwise, use the inferred run order.
variables['run_order'] = cc_info['inferred_run_order']
note("inferred run order to be: %r" % (variables['run_order'],))
if opts.verbose:
format = pprint.pformat(variables)
msg = '\n\t'.join(['using variables:'] + format.splitlines())
note(msg)
format = pprint.pformat(machine_info)
msg = '\n\t'.join(['using machine info:'] + format.splitlines())
note(msg)
format = pprint.pformat(run_info)
msg = '\n\t'.join(['using run info:'] + format.splitlines())
note(msg)
# Compute the set of flags to test.
if not opts.flags_to_test:
flags_to_test = DEFAULT_FLAGS_TO_TEST
else:
flags_to_test = [string.split(' ')
for string in opts.flags_to_test]
# Compute the set of job counts to use in full build tests.
if not opts.jobs_to_test:
jobs_to_test = [1, 2, 4, 8]
else:
jobs_to_test = opts.jobs_to_test
# Compute the build configurations to test.
if not opts.configs_to_test:
configs_to_test = ['Debug', 'Release']
else:
configs_to_test = opts.configs_to_test
# Compute the list of all tests.
all_tests = list(get_tests(opts.test_suite_externals, opts.test_subdir,
flags_to_test, jobs_to_test,
configs_to_test))
# Show the tests, if requested.
if opts.show_tests:
print >>sys.stderr, 'Available Tests'
for name in sorted(set(name for name, _ in all_tests)):
print >>sys.stderr, ' %s' % (name,)
print
raise SystemExit
# Find the tests to run.
if not opts.tests and not opts.test_filters:
tests_to_run = list(all_tests)
else:
all_test_names = set(test[0] for test in all_tests)
# Validate the test names.
requested_tests = set(opts.tests)
missing_tests = requested_tests - all_test_names
if missing_tests:
parser.error(("invalid test names %s, use --show-tests to "
"see available tests") %
(", ".join(map(repr, missing_tests)), ))
# Validate the test filters.
test_filters = [re.compile(pattern)
for pattern in opts.test_filters]
# Form the list of tests.
tests_to_run = [test
for test in all_tests
if (test[0] in requested_tests or
[True
for filter in test_filters
if filter.search(test[0])])]
if not tests_to_run:
parser.error(
"no tests requested (invalid --test or --test-filter options)!")
# Ensure output directory is available.
if not os.path.exists(g_output_dir):
os.mkdir(g_output_dir)
# Execute the run.
run_info.update(variables)
run_info['tag'] = tag = 'compile'
testsamples = []
start_time = datetime.utcnow()
g_log.info('run started')
g_log.info('using CC: %r' % opts.cc)
g_log.info('using CXX: %r' % opts.cxx)
for basename, test_fn in tests_to_run:
for success, name, samples in test_fn(basename, run_info,
variables):
g_log.info('collected samples: %r' % name)
num_samples = len(samples)
if num_samples:
samples_median = '%.4f' % (stats.median(samples),)
samples_mad = '%.4f' % (
stats.median_absolute_deviation(samples),)
else:
samples_median = samples_mad = 'N/A'
g_log.info('N=%d, median=%s, MAD=%s' % (
num_samples, samples_median, samples_mad))
test_name = '%s.%s' % (tag, name)
if not success:
testsamples.append(lnt.testing.TestSamples(
test_name + '.status',
[lnt.testing.FAIL]))
if samples:
testsamples.append(lnt.testing.TestSamples(
test_name, samples))
end_time = datetime.utcnow()
g_log.info('run complete')
# Package up the report.
machine = lnt.testing.Machine(opts.machine_name, machine_info)
run = lnt.testing.Run(start_time, end_time, info=run_info)
# Write out the report.
lnt_report_path = os.path.join(g_output_dir, 'report.json')
report = lnt.testing.Report(machine, run, testsamples)
# Save report to disk for submission.
self.print_report(report, lnt_report_path)
# Then, also print to screen if requested.
if opts.output is not None:
self.print_report(report, opts.output)
server_report = self.submit(lnt_report_path, opts)
return server_report
def __init__(self, aggregation_fn,
cur_failed, prev_failed, samples, prev_samples,
cur_hash, prev_hash, cur_profile=None, prev_profile=None,
confidence_lv=0.05, bigger_is_better=False, stable_test=True):
self.aggregation_fn = aggregation_fn
# Special case: if we're using the minimum to aggregate, swap it for max
# if bigger_is_better.
if aggregation_fn == stats.safe_min and bigger_is_better:
aggregation_fn = stats.safe_max
self.cur_hash = cur_hash
self.prev_hash = prev_hash
self.cur_profile = cur_profile
self.prev_profile = prev_profile
if samples:
self.current = aggregation_fn(samples)
else:
self.current = None
self.previous = None
# Compute the comparison status for the test value.
self.delta = 0
self.pct_delta = 0.0
if self.current and prev_samples:
prev_median = stats.median(prev_samples)
self.delta, value = absmin_diff(self.current, [prev_median])
if value != 0:
self.pct_delta = self.delta / value
self.previous = value
# If we have multiple values for this run, use that to estimate the
# distribution.
#
# We can get integer sample types here - for example if the field is
# .exec.status. Make sure we don't assert by avoiding the stats
# functions in this case.
if samples and len(samples) > 1 and isinstance(samples[0], float):
self.stddev = stats.standard_deviation(samples)
self.MAD = stats.median_absolute_deviation(samples)
self.variance = stats.variance(samples)
else:
self.stddev = None
self.MAD = None
self.variance = None
if prev_samples and len(prev_samples) > 1 and isinstance(prev_samples[0], float):
self.prev_stddev = stats.standard_deviation(prev_samples)
self.prev_MAD = stats.median_absolute_deviation(prev_samples)
self.prev_variance = stats.variance(prev_samples)
else:
self.prev_stddev = None
self.prev_MAD = None
self.prev_variance = None
self.stddev_mean = None # Only calculate this if needed.
self.failed = cur_failed
self.prev_failed = prev_failed
self.samples = samples
self.prev_samples = prev_samples
self.confidence_lv = confidence_lv
self.bigger_is_better = bigger_is_better
self.stable_test = stable_test
def v4_graph():
from lnt.server.ui import util
from lnt.testing import PASS
from lnt.util import stats
from lnt.external.stats import stats as ext_stats
ts = request.get_testsuite()
# Parse the view options.
options = {}
options['hide_lineplot'] = bool(request.args.get('hide_lineplot'))
show_lineplot = not options['hide_lineplot']
options['show_mad'] = show_mad = bool(request.args.get('show_mad'))
options['show_stddev'] = show_stddev = bool(
request.args.get('show_stddev'))
options['hide_all_points'] = hide_all_points = bool(
request.args.get('hide_all_points'))
options['show_linear_regression'] = show_linear_regression = bool(
request.args.get('show_linear_regression'))
options['show_failures'] = show_failures = bool(
request.args.get('show_failures'))
options['normalize_by_median'] = normalize_by_median = bool(
request.args.get('normalize_by_median'))
options['show_moving_average'] = moving_average = bool(
request.args.get('show_moving_average'))
options['show_moving_median'] = moving_median = bool(
request.args.get('show_moving_median'))
options['moving_window_size'] = moving_window_size = int(
request.args.get('moving_window_size', 10))
options['hide_highlight'] = bool(request.args.get('hide_highlight'))
show_highlight = not options['hide_highlight']
def convert_revision(dotted):
"""Turn a version number like 489.2.10 into something
that is ordered and sortable.
For now 489.2.10 will be returned as a tuple of ints.
"""
dotted = integral_rex.findall(dotted)
return tuple([int(d) for d in dotted])
# Load the graph parameters.
graph_parameters = []
for name, value in request.args.items():
# Plots to graph are passed as::
#
# plot.<unused>=<machine id>.<test id>.<field index>
if not name.startswith(str('plot.')):
continue
# Ignore the extra part of the key, it is unused.
machine_id_str, test_id_str, field_index_str = value.split('.')
try:
machine_id = int(machine_id_str)
test_id = int(test_id_str)
field_index = int(field_index_str)
except:
return abort(400)
if not (0 <= field_index < len(ts.sample_fields)):
return abort(404)
try:
machine = \
ts.query(ts.Machine).filter(ts.Machine.id == machine_id).one()
test = ts.query(ts.Test).filter(ts.Test.id == test_id).one()
field = ts.sample_fields[field_index]
except NoResultFound:
return abort(404)
graph_parameters.append((machine, test, field, field_index))
# Order the plots by machine name, test name and then field.
graph_parameters.sort(key=lambda (m, t, f, _): (m.name, t.name, f.name, _))
# Extract requested mean trend.
mean_parameter = None
for name, value in request.args.items():
# Mean to graph is passed as:
#
# mean=<machine id>.<field index>
if name != 'mean':
continue
machine_id_str, field_index_str = value.split('.')
try:
machine_id = int(machine_id_str)
field_index = int(field_index_str)
except ValueError:
return abort(400)
if not (0 <= field_index < len(ts.sample_fields)):
return abort(404)
try:
machine = \
ts.query(ts.Machine).filter(ts.Machine.id == machine_id).one()
except NoResultFound:
return abort(404)
field = ts.sample_fields[field_index]
mean_parameter = (machine, field)
# Sanity check the arguments.
if not graph_parameters and not mean_parameter:
return render_template("error.html", message="Nothing to graph.")
# Extract requested baselines, and their titles.
baseline_parameters = []
for name, value in request.args.items():
# Baselines to graph are passed as:
#
# baseline.title=<run id>
if not name.startswith(str('baseline.')):
continue
baseline_title = name[len('baseline.'):]
run_id_str = value
try:
run_id = int(run_id_str)
except:
return abort(400)
try:
run = ts.query(ts.Run).join(
ts.Machine).filter(ts.Run.id == run_id).one()
except:
err_msg = "The run {} was not found in the database.".format(
run_id)
return render_template("error.html", message=err_msg)
baseline_parameters.append((run, baseline_title))
# Create region of interest for run data region if we are performing a
# comparison.
revision_range = None
highlight_run_id = request.args.get('highlight_run')
if show_highlight and highlight_run_id and highlight_run_id.isdigit():
highlight_run = ts.query(
ts.Run).filter_by(id=int(highlight_run_id)).first()
if highlight_run is None:
abort(404)
# Find the neighboring runs, by order.
prev_runs = list(ts.get_previous_runs_on_machine(highlight_run, N=1))
if prev_runs:
start_rev = prev_runs[0].order.llvm_project_revision
end_rev = highlight_run.order.llvm_project_revision
revision_range = {
"start": convert_revision(start_rev),
"end": convert_revision(end_rev)
}
# Build the graph data.
legend = []
graph_plots = []
graph_datum = []
overview_plots = []
baseline_plots = []
num_plots = len(graph_parameters)
for i, (machine, test, field, field_index) in enumerate(graph_parameters):
# Determine the base plot color.
col = list(util.makeDarkColor(float(i) / num_plots))
url = "/".join([str(machine.id), str(test.id), str(field_index)])
legend.append((machine, test.name, field.name, tuple(col), url))
# Load all the field values for this test on the same machine.
#
# FIXME: Don't join to Order here, aggregate this across all the tests
# we want to load. Actually, we should just make this a single query.
#
# FIXME: Don't hard code field name.
q = ts.query(field.column, ts.Order.llvm_project_revision, ts.Run.start_time, ts.Run.id).\
join(ts.Run).join(ts.Order).\
filter(ts.Run.machine_id == machine.id).\
filter(ts.Sample.test == test).\
filter(field.column != None)
# Unless all samples requested, filter out failing tests.
if not show_failures:
if field.status_field:
q = q.filter((field.status_field.column == PASS)
| (field.status_field.column == None))
# Aggregate by revision.
data = util.multidict((rev, (val, date, run_id))
for val, rev, date, run_id in q).items()
data.sort(key=lambda sample: convert_revision(sample[0]))
graph_datum.append((test.name, data, col, field, url))
# Get baselines for this line
num_baselines = len(baseline_parameters)
for baseline_id, (baseline,
baseline_title) in enumerate(baseline_parameters):
q_baseline = ts.query(field.column, ts.Order.llvm_project_revision, ts.Run.start_time, ts.Machine.name).\
join(ts.Run).join(ts.Order).join(ts.Machine).\
filter(ts.Run.id == baseline.id).\
filter(ts.Sample.test == test).\
filter(field.column != None)
# In the event of many samples, use the mean of the samples as the baseline.
samples = []
for sample in q_baseline:
samples.append(sample[0])
# Skip this baseline if there is no data.
if not samples:
continue
mean = sum(samples) / len(samples)
# Darken the baseline color distinguish from non-baselines.
# Make a color closer to the sample than its neighbour.
color_offset = float(baseline_id) / num_baselines / 2
my_color = (i + color_offset) / num_plots
dark_col = list(util.makeDarkerColor(my_color))
str_dark_col = util.toColorString(dark_col)
baseline_plots.append({
'color': str_dark_col,
'lineWidth': 2,
'yaxis': {
'from': mean,
'to': mean
},
'name': q_baseline[0].llvm_project_revision
})
baseline_name = "Baseline {} on {}".format(baseline_title,
q_baseline[0].name)
legend.append((BaselineLegendItem(baseline_name, baseline.id),
test.name, field.name, dark_col))
# Draw mean trend if requested.
if mean_parameter:
machine, field = mean_parameter
test_name = 'Geometric Mean'
col = (0, 0, 0)
legend.append((machine, test_name, field.name, col, None))
q = ts.query(sqlalchemy.sql.func.min(field.column),
ts.Order.llvm_project_revision,
sqlalchemy.sql.func.min(ts.Run.start_time)).\
join(ts.Run).join(ts.Order).join(ts.Test).\
filter(ts.Run.machine_id == machine.id).\
filter(field.column != None).\
group_by(ts.Order.llvm_project_revision, ts.Test)
# Calculate geomean of each revision.
data = util.multidict(
((rev, date), val) for val, rev, date in q).items()
data = [(rev, [(lnt.server.reporting.analysis.calc_geomean(vals), date)
]) for ((rev, date), vals) in data]
# Sort data points according to revision number.
data.sort(key=lambda sample: convert_revision(sample[0]))
graph_datum.append((test_name, data, col, field, None))
for name, data, col, field, url in graph_datum:
# Compute the graph points.
errorbar_data = []
points_data = []
pts = []
moving_median_data = []
moving_average_data = []
if normalize_by_median:
normalize_by = 1.0 / stats.median(
[min([d[0] for d in values]) for _, values in data])
else:
normalize_by = 1.0
for pos, (point_label, datapoints) in enumerate(data):
# Get the samples.
data = [data_date[0] for data_date in datapoints]
# And the date on which they were taken.
dates = [data_date[1] for data_date in datapoints]
# Run where this point was collected.
runs = [
data_pts[2] for data_pts in datapoints if len(data_pts) == 3
]
# When we can, map x-axis to revisions, but when that is too hard
# use the position of the sample instead.
rev_x = convert_revision(point_label)
x = rev_x[0] if len(rev_x) == 1 else pos
values = [v * normalize_by for v in data]
aggregation_fn = min
if field.bigger_is_better:
aggregation_fn = max
agg_value, agg_index = \
aggregation_fn((value, index)
for (index, value) in enumerate(values))
# Generate metadata.
metadata = {"label": point_label}
metadata["date"] = str(dates[agg_index])
if runs:
metadata["runID"] = str(runs[agg_index])
if len(graph_datum) > 1:
# If there are more than one plot in the graph, also label the
# test name.
metadata["test_name"] = name
pts.append((x, agg_value, metadata))
# Add the individual points, if requested.
# For each point add a text label for the mouse over.
if not hide_all_points:
for i, v in enumerate(values):
point_metadata = dict(metadata)
point_metadata["date"] = str(dates[i])
points_data.append((x, v, point_metadata))
# Add the standard deviation error bar, if requested.
if show_stddev:
mean = stats.mean(values)
sigma = stats.standard_deviation(values)
errorbar_data.append((x, mean, sigma))
# Add the MAD error bar, if requested.
if show_mad:
med = stats.median(values)
mad = stats.median_absolute_deviation(values, med)
errorbar_data.append((x, med, mad))
# Compute the moving average and or moving median of our data if requested.
if moving_average or moving_median:
fun = None
def compute_moving_average(x, window, average_list, median_list):
average_list.append((x, lnt.util.stats.mean(window)))
def compute_moving_median(x, window, average_list, median_list):
median_list.append((x, lnt.util.stats.median(window)))
def compute_moving_average_and_median(x, window, average_list,
median_list):
average_list.append((x, lnt.util.stats.mean(window)))
median_list.append((x, lnt.util.stats.median(window)))
if moving_average and moving_median:
fun = compute_moving_average_and_median
elif moving_average:
fun = compute_moving_average
else:
fun = compute_moving_median
len_pts = len(pts)
for i in range(len_pts):
start_index = max(0, i - moving_window_size)
end_index = min(len_pts, i + moving_window_size)
window_pts = [x[1] for x in pts[start_index:end_index]]
fun(pts[i][0], window_pts, moving_average_data,
moving_median_data)
# On the overview, we always show the line plot.
overview_plots.append({"data": pts, "color": util.toColorString(col)})
# Add the minimum line plot, if requested.
if show_lineplot:
plot = {"data": pts, "color": util.toColorString(col)}
if url:
plot["url"] = url
graph_plots.append(plot)
# Add regression line, if requested.
if show_linear_regression:
xs = [t for t, v, _ in pts]
ys = [v for t, v, _ in pts]
# We compute the regression line in terms of a normalized X scale.
x_min, x_max = min(xs), max(xs)
try:
norm_xs = [(x - x_min) / (x_max - x_min) for x in xs]
except ZeroDivisionError:
norm_xs = xs
try:
info = ext_stats.linregress(norm_xs, ys)
except ZeroDivisionError:
info = None
except ValueError:
info = None
if info is not None:
slope, intercept, _, _, _ = info
reglin_col = [c * .7 for c in col]
reglin_pts = [(x_min, 0.0 * slope + intercept),
(x_max, 1.0 * slope + intercept)]
graph_plots.insert(
0, {
"data": reglin_pts,
"color": util.toColorString(reglin_col),
"lines": {
"lineWidth": 2
},
"shadowSize": 4
})
# Add the points plot, if used.
if points_data:
pts_col = (0, 0, 0)
plot = {
"data": points_data,
"color": util.toColorString(pts_col),
"lines": {
"show": False
},
"points": {
"show": True,
"radius": .25,
"fill": True
}
}
if url:
plot['url'] = url
graph_plots.append(plot)
# Add the error bar plot, if used.
if errorbar_data:
bar_col = [c * .7 for c in col]
graph_plots.append({
"data": errorbar_data,
"lines": {
"show": False
},
"color": util.toColorString(bar_col),
"points": {
"errorbars": "y",
"yerr": {
"show": True,
"lowerCap": "-",
"upperCap": "-",
"lineWidth": 1
}
}
})
# Add the moving average plot, if used.
if moving_average_data:
col = [0.32, 0.6, 0.0]
graph_plots.append({
"data": moving_average_data,
"color": util.toColorString(col)
})
# Add the moving median plot, if used.
if moving_median_data:
col = [0.75, 0.0, 1.0]
graph_plots.append({
"data": moving_median_data,
"color": util.toColorString(col)
})
if bool(request.args.get('json')):
json_obj = dict()
json_obj['data'] = graph_plots
# Flatten ORM machine objects to their string names.
simple_type_legend = []
for machine, test, unit, color, url in legend:
# Flatten name, make color a dict.
new_entry = {
'name': machine.name,
'test': test,
'unit': unit,
'color': util.toColorString(color),
'url': url
}
simple_type_legend.append(new_entry)
json_obj['legend'] = simple_type_legend
json_obj['revision_range'] = revision_range
json_obj['current_options'] = options
json_obj['test_suite_name'] = ts.name
json_obj['baselines'] = baseline_plots
return flask.jsonify(**json_obj)
return render_template("v4_graph.html",
ts=ts,
options=options,
revision_range=revision_range,
graph_plots=graph_plots,
overview_plots=overview_plots,
legend=legend,
baseline_plots=baseline_plots)