#!/usr/bin/python3
#
# Script for comparing buildstats from two different builds
#
# Copyright (c) 2016, Intel Corporation.
#
# This program is free software; you can redistribute it and/or modify it
# under the terms and conditions of the GNU General Public License,
# version 2, as published by the Free Software Foundation.
#
# This program is distributed in the hope it will be useful, but WITHOUT
# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
# FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
# more details.
#
import argparse
import glob
import logging
import math
import os
import sys
from operator import attrgetter
# Import oe libs
scripts_path = os.path.dirname(os.path.realpath(__file__))
sys.path.append(os.path.join(scripts_path, 'lib'))
from buildstats import BuildStats, diff_buildstats, taskdiff_fields, BSVerDiff
# Setup logging
logging.basicConfig(level=logging.INFO, format="%(levelname)s: %(message)s")
log = logging.getLogger()
class ScriptError(Exception):
"""Exception for internal error handling of this script"""
pass
def read_buildstats(path, multi):
"""Read buildstats"""
if not os.path.exists(path):
raise ScriptError("No such file or directory: {}".format(path))
if os.path.isfile(path):
return BuildStats.from_file_json(path)
if os.path.isfile(os.path.join(path, 'build_stats')):
return BuildStats.from_dir(path)
# Handle a non-buildstat directory
subpaths = sorted(glob.glob(path + '/*'))
if len(subpaths) > 1:
if multi:
log.info("Averaging over {} buildstats from {}".format(
len(subpaths), path))
else:
raise ScriptError("Multiple buildstats found in '{}'. Please give "
"a single buildstat directory of use the --multi "
"option".format(path))
bs = None
for subpath in subpaths:
if os.path.isfile(subpath):
_bs = BuildStats.from_file_json(subpath)
else:
_bs = BuildStats.from_dir(subpath)
if bs is None:
bs = _bs
else:
bs.aggregate(_bs)
if not bs:
raise ScriptError("No buildstats found under {}".format(path))
return bs
def print_ver_diff(bs1, bs2):
"""Print package version differences"""
diff = BSVerDiff(bs1, bs2)
maxlen = max([len(r) for r in set(bs1.keys()).union(set(bs2.keys()))])
fmt_str = " {:{maxlen}} ({})"
if diff.new:
print("\nNEW RECIPES:")
print("------------")
for name, val in sorted(diff.new.items()):
print(fmt_str.format(name, val.nevr, maxlen=maxlen))
if diff.dropped:
print("\nDROPPED RECIPES:")
print("----------------")
for name, val in sorted(diff.dropped.items()):
print(fmt_str.format(name, val.nevr, maxlen=maxlen))
fmt_str = " {0:{maxlen}} {1:<20} ({2})"
if diff.rchanged:
print("\nREVISION CHANGED:")
print("-----------------")
for name, val in sorted(diff.rchanged.items()):
field1 = "{} -> {}".format(val.left.revision, val.right.revision)
field2 = "{} -> {}".format(val.left.nevr, val.right.nevr)
print(fmt_str.format(name, field1, field2, maxlen=maxlen))
if diff.vchanged:
print("\nVERSION CHANGED:")
print("----------------")
for name, val in sorted(diff.vchanged.items()):
field1 = "{} -> {}".format(val.left.version, val.right.version)
field2 = "{} -> {}".format(val.left.nevr, val.right.nevr)
print(fmt_str.format(name, field1, field2, maxlen=maxlen))
if diff.echanged:
print("\nEPOCH CHANGED:")
print("--------------")
for name, val in sorted(diff.echanged.items()):
field1 = "{} -> {}".format(val.left.epoch, val.right.epoch)
field2 = "{} -> {}".format(val.left.nevr, val.right.nevr)
print(fmt_str.format(name, field1, field2, maxlen=maxlen))
def print_task_diff(bs1, bs2, val_type, min_val=0, min_absdiff=0, sort_by=('absdiff',)):
"""Diff task execution times"""
def val_to_str(val, human_readable=False):
"""Convert raw value to printable string"""
def hms_time(secs):
"""Get time in human-readable HH:MM:SS format"""
h = int(secs / 3600)
m = int((secs % 3600) / 60)
s = secs % 60
if h == 0:
return "{:02d}:{:04.1f}".format(m, s)
else:
return "{:d}:{:02d}:{:04.1f}".format(h, m, s)
if 'time' in val_type:
if human_readable:
return hms_time(val)
else:
return "{:.1f}s".format(val)
elif 'bytes' in val_type and human_readable:
prefix = ['', 'Ki', 'Mi', 'Gi', 'Ti', 'Pi']
dec = int(math.log(val, 2) / 10)
prec = 1 if dec > 0 else 0
return "{:.{prec}f}{}B".format(val / (2 ** (10 * dec)),
prefix[dec], prec=prec)
elif 'ops' in val_type and human_readable:
prefix = ['', 'k', 'M', 'G', 'T', 'P']
dec = int(math.log(val, 1000))
prec = 1 if dec > 0 else 0
return "{:.{prec}f}{}ops".format(val / (1000 ** dec),
prefix[dec], prec=prec)
return str(int(val))
def sum_vals(buildstats):
"""Get cumulative sum of all tasks"""
total = 0.0
for recipe_data in buildstats.values():
for bs_task in recipe_data.tasks.values():
total += getattr(bs_task, val_type)
return total
if min_val:
print("Ignoring tasks less than {} ({})".format(
val_to_str(min_val, True), val_to_str(min_val)))
if min_absdiff:
print("Ignoring differences less than {} ({})".format(
val_to_str(min_absdiff, True), val_to_str(min_absdiff)))
# Prepare the data
tasks_diff = diff_buildstats(bs1, bs2, val_type, min_val, min_absdiff)
# Sort our list
for field in reversed(sort_by):
if field.startswith('-'):
field = field[1:]
reverse = True
else:
reverse = False
tasks_diff = sorted(tasks_diff, key=attrgetter(field), reverse=reverse)
linedata = [(' ', 'PKG', ' ', 'TASK', 'ABSDIFF', 'RELDIFF',
val_type.upper() + '1', val_type.upper() + '2')]
field_lens = dict([('len_{}'.format(i), len(f)) for i, f in enumerate(linedata[0])])
# Prepare fields in string format and measure field lengths
for diff in tasks_diff:
task_prefix = diff.task_op if diff.pkg_op == ' ' else ' '
linedata.append((diff.pkg_op, diff.pkg, task_prefix, diff.task,
val_to_str(diff.absdiff),
'{:+.1f}%'.format(diff.reldiff),
val_to_str(diff.value1),
val_to_str(diff.value2)))
for i, field in enumerate(linedata[-1]):
key = 'len_{}'.format(i)
if len(field) > field_lens[key]:
field_lens[key] = len(field)
# Print data
print()
for fields in linedata:
print("{:{len_0}}{:{len_1}} {:{len_2}}{:{len_3}} {:>{len_4}} {:>{len_5}} {:>{len_6}} -> {:{len_7}}".format(
*fields, **field_lens))
# Print summary of the diffs
total1 = sum_vals(bs1)
total2 = sum_vals(bs2)
print("\nCumulative {}:".format(val_type))
print (" {} {:+.1f}% {} ({}) -> {} ({})".format(
val_to_str(total2 - total1), 100 * (total2-total1) / total1,
val_to_str(total1, True), val_to_str(total1),
val_to_str(total2, True), val_to_str(total2)))
def parse_args(argv):
"""Parse cmdline arguments"""
description="""
Script for comparing buildstats of two separate builds."""
parser = argparse.ArgumentParser(
formatter_class=argparse.ArgumentDefaultsHelpFormatter,
description=description)
min_val_defaults = {'cputime': 3.0,
'read_bytes': 524288,
'write_bytes': 524288,
'read_ops': 500,
'write_ops': 500,
'walltime': 5}
min_absdiff_defaults = {'cputime': 1.0,
'read_bytes': 131072,
'write_bytes': 131072,
'read_ops': 50,
'write_ops': 50,
'walltime': 2}
parser.add_argument('--debug', '-d', action='store_true',
help="Verbose logging")
parser.add_argument('--ver-diff', action='store_true',
help="Show package version differences and exit")
parser.add_argument('--diff-attr', default='cputime',
choices=min_val_defaults.keys(),
help="Buildstat attribute which to compare")
parser.add_argument('--min-val', default=min_val_defaults, type=float,
help="Filter out tasks less than MIN_VAL. "
"Default depends on --diff-attr.")
parser.add_argument('--min-absdiff', default=min_absdiff_defaults, type=float,
help="Filter out tasks whose difference is less than "
"MIN_ABSDIFF, Default depends on --diff-attr.")
parser.add_argument('--sort-by', default='absdiff',
help="Comma-separated list of field sort order. "
"Prepend the field name with '-' for reversed sort. "
"Available fields are: {}".format(', '.join(taskdiff_fields)))
parser.add_argument('--multi', action='store_true',
help="Read all buildstats from the given paths and "
"average over them")
parser.add_argument('buildstats1', metavar='BUILDSTATS1', help="'Left' buildstat")
parser.add_argument('buildstats2', metavar='BUILDSTATS2', help="'Right' buildstat")
args = parser.parse_args(argv)
# We do not nedd/want to read all buildstats if we just want to look at the
# package versions
if args.ver_diff:
args.multi = False
# Handle defaults for the filter arguments
if args.min_val is min_val_defaults:
args.min_val = min_val_defaults[args.diff_attr]
if args.min_absdiff is min_absdiff_defaults:
args.min_absdiff = min_absdiff_defaults[args.diff_attr]
return args
def main(argv=None):
"""Script entry point"""
args = parse_args(argv)
if args.debug:
log.setLevel(logging.DEBUG)
# Validate sort fields
sort_by = []
for field in args.sort_by.split(','):
if field.lstrip('-') not in taskdiff_fields:
log.error("Invalid sort field '%s' (must be one of: %s)" %
(field, ', '.join(taskdiff_fields)))
sys.exit(1)
sort_by.append(field)
try:
bs1 = read_buildstats(args.buildstats1, args.multi)
bs2 = read_buildstats(args.buildstats2, args.multi)
if args.ver_diff:
print_ver_diff(bs1, bs2)
else:
print_task_diff(bs1, bs2, args.diff_attr, args.min_val,
args.min_absdiff, sort_by)
except ScriptError as err:
log.error(str(err))
return 1
return 0
if __name__ == "__main__":
sys.exit(main())