Protocol Buffers - Google's data interchange format (grpc依赖)
https://developers.google.com/protocol-buffers/
You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
88 lines
3.1 KiB
88 lines
3.1 KiB
#!/usr/bin/env python |
|
# Protocol Buffers - Google's data interchange format |
|
# Copyright 2008 Google Inc. All rights reserved. |
|
# |
|
# Use of this source code is governed by a BSD-style |
|
# license that can be found in the LICENSE file or at |
|
# https://developers.google.com/open-source/licenses/bsd |
|
|
|
"""Script to update a failure list file to add/remove failures. |
|
|
|
When adding, will attempt to place them in their correct lexicographical |
|
position relative to other test names. This requires that the failure list is |
|
already sorted. This does not guarantee that the tests will appear neatly one |
|
after the other, as there may be comments in between. If the failure list |
|
is not sorted, sorted blocks may be produced, but the list as a whole will not. |
|
Lastly, be wary of lists that have tests stripped from OSS; if you catch that |
|
a test was added inside a stripped block, you may need to move it out. |
|
|
|
This is sort of like comm(1), except it recognizes comments and ignores them. |
|
""" |
|
|
|
import argparse |
|
|
|
parser = argparse.ArgumentParser( |
|
description='Adds/removes failures from the failure list.') |
|
parser.add_argument('filename', type=str, help='failure list file to update') |
|
parser.add_argument('--add', dest='add_list', action='append') |
|
parser.add_argument('--remove', dest='remove_list', action='append') |
|
|
|
DEFAULT_ALIGNMENT = 114 |
|
args = parser.parse_args() |
|
|
|
add_set = set() |
|
remove_set = set() |
|
|
|
# Adds test + failure message |
|
for add_file in (args.add_list or []): |
|
with open(add_file) as f: |
|
for line in f: |
|
add_set.add(line) |
|
|
|
# We only remove by name |
|
for remove_file in (args.remove_list or []): |
|
with open(remove_file) as f: |
|
for line in f: |
|
if line in add_set: |
|
raise Exception("Asked to both add and remove test: " + line) |
|
remove_set.add(line.split('#')[0].strip()) |
|
|
|
add_list = sorted(add_set, reverse=True) |
|
|
|
with open(args.filename) as in_file: |
|
existing_list = in_file.read() |
|
|
|
with open(args.filename, 'w') as f: |
|
split_lines = existing_list.splitlines(True) |
|
# Make sure that we have one newline for potential tests that have to be |
|
# added at the end. |
|
split_lines[-1] = split_lines[-1].rstrip() + '\n' |
|
for line in split_lines: |
|
test = line.split('#')[0].strip() |
|
# As long as the tests we are adding appear lexicographically before our |
|
# read test, put them first followed by our read test. |
|
while add_list and test > add_list[-1]: |
|
f.write(add_list.pop()) |
|
if test not in remove_set: |
|
f.write(line) |
|
# Any remaining tests are added at the end |
|
while add_list: |
|
f.write(add_list.pop()) |
|
|
|
# Update our read of the existing file |
|
with open(args.filename, 'r') as f: |
|
existing_list = f.read() |
|
|
|
# Actual alignment of failure messages to 'DEFAULT_ALIGNMENT' |
|
# If test name exceeds DEFAULT_ALIGNMENT, it cannot and will not be aligned. |
|
with open(args.filename, 'w') as f: |
|
for line in existing_list.splitlines(True): |
|
split = line.split('#', 1) |
|
test = split[0].strip() |
|
if len(split) > 1 and test: |
|
message = split[1].lstrip() |
|
spaces = ' ' * (DEFAULT_ALIGNMENT - len(test)) |
|
line = test + spaces + ' # ' + message |
|
f.write(line) |
|
else: # ignore blank lines/lines that do not have '#'/comments |
|
f.write(line)
|
|
|