aplanas/git-importer
1
0
Fork 0
forked from importers/git-importer
Code Issues Pull Requests Packages Projects Releases Wiki Activity
bdc1dc0cc9
git-importer/git-importer.py
Stephan Kulow bdc1dc0cc9 Split out History class
2022-10-17 15:02:34 +02:00

710 lines
25 KiB
Python
Executable File
Raw Blame History

#!/usr/bin/python3
import argparse
import errno
import functools
import hashlib
import logging
import pathlib
import shutil
import sys
import time
import urllib.parse
import xml.etree.ElementTree as ET
from urllib.error import HTTPError
import osc.core
import requests
from request import Request
from git import Git
from history import History
from binary import is_binary_or_large
# Add a retry wrapper for some of the HTTP actions.
def retry(func):
def wrapper(*args, **kwargs):
retry = 0
while retry < 5:
try:
return func(*args, **kwargs)
except HTTPError as e:
if 500 <= e.code <= 599:
retry += 1
logging.warning(
f"HTTPError {e.code} -- Retrying {args[0]} ({retry})"
)
# TODO: remove when move to async
time.sleep(0.5)
else:
raise
except urllib.error.URLError as e:
if e.reason.errno in (errno.ENETUNREACH, errno.EADDRNOTAVAIL):
retry += 1
logging.warning(f"URLError {e} -- Retrying {args[0]} ({retry})")
time.sleep(0.5)
else:
logging.warning(f"URLError {e.errno} uncaught")
raise
except OSError as e:
if e.errno in (errno.ENETUNREACH, errno.EADDRNOTAVAIL): # sporadically hits cloud VMs :(
retry += 1
logging.warning(f"OSError {e} -- Retrying {args[0]} ({retry})")
# TODO: remove when move to async
time.sleep(0.5)
else:
logging.warning(f"OSError {e.errno} uncaught")
raise
return wrapper
osc.core.http_GET = retry(osc.core.http_GET)
URL_OBS = "https://api.opensuse.org"
URL_IBS = "https://api.suse.de"
# The order is relevant (from older to newer initial codebase)
PROJECTS = [
("openSUSE:Factory", "factory", URL_OBS),
# ("SUSE:SLE-12:GA", "SLE_12", URL_IBS),
# ("SUSE:SLE-12:Update", "SLE_12", URL_IBS),
# ("SUSE:SLE-12-SP1:GA", "SLE_12_SP1", URL_IBS),
# ("SUSE:SLE-12-SP1:Update", "SLE_12_SP1", URL_IBS),
# ("SUSE:SLE-12-SP2:GA", "SLE_12_SP2", URL_IBS),
# ("SUSE:SLE-12-SP2:Update", "SLE_12_SP2", URL_IBS),
# ("SUSE:SLE-12-SP3:GA", "SLE_12_SP3", URL_IBS),
# ("SUSE:SLE-12-SP3:Update", "SLE_12_SP3", URL_IBS),
# ("SUSE:SLE-12-SP4:GA", "SLE_12_SP4", URL_IBS),
# ("SUSE:SLE-12-SP4:Update", "SLE_12_SP4", URL_IBS),
# ("SUSE:SLE-12-SP5:GA", "SLE_12_SP5", URL_IBS),
# ("SUSE:SLE-12-SP5:Update", "SLE_12_SP5", URL_IBS),
# ("SUSE:SLE-15:GA", "SLE_15", URL_IBS),
# ("SUSE:SLE-15:Update", "SLE_15", URL_IBS),
# ("SUSE:SLE-15-SP1:GA", "SLE_15_SP1", URL_IBS),
# ("SUSE:SLE-15-SP1:Update", "SLE_15_SP1", URL_IBS),
# ("SUSE:SLE-15-SP2:GA", "SLE_15_SP2", URL_IBS),
# ("SUSE:SLE-15-SP2:Update", "SLE_15_SP2", URL_IBS),
# ("SUSE:SLE-15-SP3:GA", "SLE_15_SP3", URL_IBS),
# ("SUSE:SLE-15-SP3:Update", "SLE_15_SP3", URL_IBS),
# ("SUSE:SLE-15-SP4:GA", "SLE_15_SP4", URL_IBS),
# ("SUSE:SLE-15-SP4:Update", "SLE_15_SP4", URL_IBS),
]
def _hash(hash_alg, file_or_path):
h = hash_alg()
def __hash(f):
while chunk := f.read(1024 * 4):
h.update(chunk)
if hasattr(file_or_path, "read"):
__hash(file_or_path)
else:
with file_or_path.open("rb") as f:
__hash(f)
return h.hexdigest()
md5 = functools.partial(_hash, hashlib.md5)
sha256 = functools.partial(_hash, hashlib.sha256)
def _files_hash(hash_alg, dirpath):
"""List of (filepath, md5) for a directory"""
# TODO: do it async or multythread
files = [f for f in dirpath.iterdir() if f.is_file()]
return [(f.parts[-1], hash_alg(f)) for f in files]
files_md5 = functools.partial(_files_hash, md5)
files_sha256 = functools.partial(_files_hash, sha256)
class OBS:
def __init__(self, url=None):
if url:
self.change_url(url)
def change_url(self, url):
self.url = url
osc.conf.get_config(override_apiurl=url)
def _xml(self, url_path, **params):
url = osc.core.makeurl(self.url, [url_path], params)
logging.debug(f"GET {url}")
return ET.parse(osc.core.http_GET(url)).getroot()
def _meta(self, project, package, **params):
try:
root = self._xml(f"source/{project}/{package}/_meta", **params)
except HTTPError:
logging.error(f"Package [{project}/{package} {params}] has no meta")
return None
return root
def _history(self, project, package, **params):
try:
root = self._xml(f"source/{project}/{package}/_history", **params)
except HTTPError:
logging.error(f"Package [{project}/{package} {params}] has no history")
return None
return root
def _link(self, project, package, rev):
try:
root = self._xml(f"source/{project}/{package}/_link", rev=rev)
except HTTPError:
logging.info("Package has no link")
return None
except ET.ParseError:
logging.error(
f"Package [{project}/{package} rev={rev}] _link can't be parsed"
)
return root
def _request(self, requestid):
try:
root = self._xml(f"request/{requestid}")
except HTTPError:
logging.warning(f"Cannot fetch request {requestid}")
return None
return root
def exists(self, project, package):
root = self._meta(project, package)
if root is None:
return False
return root.get("project") == project
def devel_project(self, project, package):
root = self._meta(project, package)
devel = root.find("devel")
if devel is None:
return None
return devel.get("project")
def request(self, requestid):
root = self._request(requestid)
if root is not None:
return Request().parse(root)
def files(self, project, package, revision):
root = self._xml(f"source/{project}/{package}", rev=revision, expand=1)
return [
(e.get("name"), int(e.get("size")), e.get("md5"))
for e in root.findall("entry")
]
def _download(self, project, package, name, revision):
url = osc.core.makeurl(
self.url,
["source", project, package, urllib.parse.quote(name)],
{"rev": revision, "expand": 1},
)
return osc.core.http_GET(url)
def download(self, project, package, name, revision, dirpath):
with (dirpath / name).open("wb") as f:
f.write(self._download(project, package, name, revision).read())
class ProxySHA256:
def __init__(self, obs, url=None, enabled=True):
self.obs = obs
self.url = url if url else "http://source.dyn.cloud.suse.de"
self.enabled = enabled
self.hashes = None
self.texts = set()
def load_package(self, package):
# _project is unreachable for the proxy - due to being a fake package
if package == "_project":
self.enabled = False
self.texts = set(["_config", "_service"])
self.hashes = dict()
return
logging.info("Retrieve all previously defined SHA256")
response = requests.get(f"http://source.dyn.cloud.suse.de/package/{package}")
if response.status_code == 200:
json = response.json()
self.hashes = json["shas"]
self.texts = set(json["texts"])
def get(self, package, name, file_md5):
key = f"{file_md5}-{name}"
if self.hashes is None:
if self.enabled:
self.load_package(package)
else:
self.hashes = {}
return self.hashes.get(key, None)
def _proxy_put(self, project, package, name, revision, file_md5, size):
quoted_name = urllib.parse.quote(name)
url = f"{self.obs.url}/public/source/{project}/{package}/{quoted_name}?rev={revision}"
response = requests.put(
self.url,
data={
"hash": file_md5,
"filename": name,
"url": url,
"package": package,
},
)
if response.status_code != 200:
raise Exception(f"Redirector error on {self.url} for {url}")
key = (file_md5, name)
self.hashes[key] = {
"sha256": response.content.decode("utf-8"),
"fsize": size,
}
return self.hashes[key]
def _obs_put(self, project, package, name, revision, file_md5, size):
key = (file_md5, name)
self.hashes[key] = {
"sha256": sha256(self.obs._download(project, package, name, revision)),
"fsize": size,
}
return self.hashes[key]
def put(self, project, package, name, revision, file_md5, size):
if not self.enabled:
return self._obs_put(project, package, name, revision, file_md5, size)
return self._proxy_put(project, package, name, revision, file_md5, size)
def is_text(self, filename):
return filename in self.texts
def get_or_put(self, project, package, name, revision, file_md5, size):
result = self.get(package, name, file_md5)
if not result:
result = self.put(project, package, name, revision, file_md5, size)
# Sanity check
if result["fsize"] != size:
raise Exception(f"Redirector has different size for {name}")
return result
class Importer:
def __init__(self, projects, package, repodir, search_ancestor, rebase_devel):
# The idea is to create each commit in order, and draw the
# same graph described by the revisions timeline. For that we
# need first to fetch all the revisions and sort them
# linearly, based on the timestamp.
#
# After that we recreate the commits, and if one revision is a
# request that contains a target inside the projects in the
# "history", we create a merge commit.
#
# Optionally, if a flag is set, we will try to find a common
# "Initial commit" from a reference branch (the first one in
# "projects", that is safe to assume to be "openSUSE:Factory".
# This is not always a good idea. For example, in a normal
# situation the "devel" project history is older than
# "factory", and we can root the tree on it. But for some
# other projects we lost partially the "devel" history project
# (could be moved), and "factory" is not the root.
self.package = package
self.search_ancestor = search_ancestor
self.rebase_devel = rebase_devel
self.obs = OBS()
self.git = Git(
repodir,
committer="Git OBS Bridge",
committer_email="obsbridge@suse.de",
).create()
self.proxy_sha256 = ProxySHA256(self.obs, enabled=True)
self.history = History(self.obs, self.package)
# Add the "devel" project
(project, branch, api_url) = projects[0]
assert project == "openSUSE:Factory"
self.obs.change_url(api_url)
devel_project = self.obs.devel_project(project, package)
if devel_project:
self.projects = [(devel_project, "devel", api_url)] + projects
else:
self.projects = projects
# Associate the branch and api_url information per project
self.projects_info = {
project: (branch, api_url) for (project, branch, api_url) in self.projects
}
def download(self, revision):
obs_files = self.obs.files(revision.project, revision.package, revision.srcmd5)
git_files = {
(f.name, f.stat().st_size, md5(f))
for f in self.git.path.iterdir()
if f.is_file() and f.name not in (".gitattributes")
}
# Overwrite ".gitattributes" with the
self.git.add_default_lfs_gitattributes(force=True)
# Download each file in OBS if it is not a binary (or large)
# file
for (name, size, file_md5) in obs_files:
# this file creates easily 100k commits and is just useless data :(
# unfortunately it's stored in the same meta package as the project config
if revision.package == "_project" and name == "_staging_workflow":
continue
# have such files been detected as text mimetype before?
is_text = self.proxy_sha256.is_text(name)
if not is_text and is_binary_or_large(name, size):
file_sha256 = self.proxy_sha256.get_or_put(
revision.project,
revision.package,
name,
revision.srcmd5,
file_md5,
size,
)
self.git.add_lfs(name, file_sha256["sha256"], size)
else:
if (name, size, file_md5) not in git_files:
print(f"Download {name}")
self.obs.download(
revision.project,
revision.package,
name,
revision.srcmd5,
self.git.path,
)
# Validate the MD5 of the downloaded file
if md5(self.git.path / name) != file_md5:
raise Exception(f"Download error in {name}")
self.git.add(name)
# Remove extra files
obs_names = {n for (n, _, _) in obs_files}
git_names = {n for (n, _, _) in git_files}
for name in git_names - obs_names:
print(f"Remove {name}")
self.git.remove(name)
def import_all_revisions(self, gc):
# Fetch all the requests and sort them. Ideally we should
# build the graph here, to avoid new commits before the merge.
# For now we will sort them and invalidate the commits if
# "rebase_devel" is set.
self.history.fetch_all_revisions(self.projects)
revisions = self.history.sort_all_revisions()
logging.debug(f"Selected import order for {self.package}")
for revision in revisions:
logging.debug(revision)
gc_cnt = gc
for revision in revisions:
gc_cnt -= 1
if gc_cnt <= 0 and gc:
self.git.gc()
gc_cnt = gc
self.import_revision(revision)
def import_new_revision_with_request(self, revision, request):
"""Create a new branch as a result of a merge"""
submitted_revision = self.history.find_revision(
request.source, request.revisionid, revision.time
)
if not submitted_revision:
logging.warning(f"Request {request} does not connect to a known revision")
return False
if not submitted_revision.commit:
# If the revision appointed by the request is not part of
# the git history, we can have an ordering problem. One
# example is "premake4".
self.import_revision(submitted_revision)
assert submitted_revision.commit is not None
project = revision.project
branch, _ = self.projects_info[project]
# TODO: add an empty commit marking the acceptenace of the request (see discussion in PR 2858)
self.git.branch(branch, submitted_revision.commit)
self.git.clean()
self.git.checkout(branch)
logging.info(f"Create new branch based on {submitted_revision.commit}")
revision.commit = submitted_revision.commit
def _rebase_branch_history(self, project, revision):
branch, _ = self.projects_info[project]
history = self.history[project]
revision_index = history.index(revision)
for index in range(revision_index + 1, len(history)):
revision = history[index]
# We are done when we have one non-commited revision
if not revision.commit:
return
logging.info(f"Rebasing {revision} from {branch}")
revision.commit = None
self.import_revision(revision)
def import_revision_with_request(self, revision, request):
"""Import a single revision via a merge"""
submitted_revision = self.history.find_revision(
request.source, request.revisionid, revision.time
)
if not submitted_revision:
logging.warning(f"Request {request} does not connect to a known revision")
return False
assert submitted_revision.commit is not None
# TODO: detect a revision, case in point
# Base:System/bash/284 -> rq683701 -> accept O:F/151
# -> autocommit Base:System/bash/285
# Revert lead to openSUSE:Factory/bash/152
# Base:System/286 restored the reverted code in devel project
# rq684575 was created and accepted as O:F/153
# But the 284-285 and the 285-286 changeset is seen as empty
# as the revert was never in Base:System, so the
# submitted_revision of 684575 has no commit
if submitted_revision.commit == "EMPTY":
logging.warning("Empty commit submitted?!")
return False
message = (
f"Accepting request {revision.requestid}: {revision.comment}\n\n{revision}"
)
commit = self.git.merge(
# TODO: revision.userid or request.creator?
f"OBS User {revision.userid}",
"null@suse.de",
revision.time,
message,
submitted_revision.commit,
)
if commit == "EMPTY":
logging.warning("Empty merge. Ignoring the revision and the request")
self.git.merge_abort()
revision.commit = commit
return False
if commit == "CONFLICT":
logging.info("Merge conflict. Downloading revision")
self.download(revision)
message = f"CONFLICT {message}"
commit = self.git.merge(
f"OBS User {revision.userid}",
"null@suse.de",
revision.time,
message,
submitted_revision.commit,
merged=True,
)
assert commit and commit != "CONFLICT"
logging.info(f"Merge with {submitted_revision.commit} into {commit}")
revision.commit = commit
# TODO: There are more checks to do, like for example, the
# last commit into the non-devel branch should be a merge from
# the devel branch
if self.rebase_devel:
branch, _ = self.projects_info.get(request.source, (None, None))
if branch == "devel":
self.git.repo.references[f"refs/heads/{branch}"].set_target(commit)
self._rebase_branch_history(request.source, submitted_revision)
return True
def matching_request(self, revision):
request = self.obs.request(revision.requestid)
if not request:
return None
# to be handled by the caller
if request.type() != "submit":
return request
if request.source not in self.projects_info:
logging.info("Request from a non exported project")
return None
if request.target != revision.project:
# This seems to happen when the devel project gets
# reinitialized (for example, SR#943593 in 7zip, or
# SR#437901 in ColorFull)
logging.info("Request target different from current project")
return None
if request.source == request.target:
# this is not a merge, but a different way to do a
# contribution to the (devel) project - see bindfs's rev 1
logging.info("Request within the same project")
return None
return request
def import_revision(self, revision):
"""Import a single revision into git"""
project = revision.project
branch, api_url = self.projects_info[project]
logging.info(f"Importing [{revision}] to {branch}")
self.obs.change_url(api_url)
# Populate linkrev and replace srcmd5 from the linked
# revision. If the expansion fails, the revision will be ignored
# and not imported.
if not revision.check_expanded():
logging.warning(f"Broken revision")
revision.ignored = True
return
# When doing a SR, we see also a revision in the origin
# project with the outgoing request, but without changes in
# the project. We can ignore them.
#
# If there is a request ID, it will be filtered out later,
# when the target project is different from itself.
if revision.userid == "autobuild" and not revision.requestid:
logging.info("Ignoring autocommit")
revision.ignored = True
return
if revision.userid == "buildservice-autocommit":
logging.info("Ignoring autocommit")
revision.ignored = True
return
# Create the reference if the branch is new. If so return
# True.
new_branch = self.git.checkout(branch)
if revision.requestid:
request = self.matching_request(revision)
if request:
if request.type() == "delete":
# TODO: after this comes a restore, this should be collapsed
# before even hitting git
logging.info("Delete request ignored")
revision.ignored = True
return
logging.debug(f"Found matching request: #{revision.project} #{request}")
if new_branch:
self.import_new_revision_with_request(revision, request)
return
if self.import_revision_with_request(revision, request):
return
# Import revision as a single commit (without merging)
self.download(revision)
if new_branch or self.git.is_dirty():
commit = self.git.commit(
f"OBS User {revision.userid}",
"null@suse.de",
revision.time,
# TODO: Normalize better the commit message
f"{revision.comment}\n\n{revision}",
# Create an empty commit only if is a new branch
allow_empty=new_branch,
)
revision.commit = commit
logging.info(f"Commit {commit}")
else:
logging.info("Skip empty commit")
revision.ignored = True
def main():
parser = argparse.ArgumentParser(description="OBS history importer into git")
parser.add_argument("package", help="OBS package name")
parser.add_argument(
"-r",
"--repodir",
required=False,
type=pathlib.Path,
help="Local git repository directory",
)
parser.add_argument(
"-f",
"--force",
action="store_true",
help="If the repository directory exists, remove it",
)
parser.add_argument(
"-a",
"--search-ancestor",
action="store_true",
help="Search closest ancestor candidate for initial commit",
)
parser.add_argument(
"-d",
"--rebase-devel",
action="store_true",
help="The devel project with be rebased after a merge",
)
parser.add_argument(
"-g",
"--gc",
metavar="N",
type=int,
default=200,
help="Garbage recollect and pack the git history each N commits",
)
parser.add_argument(
"--level",
"-l",
default="INFO",
help="logging level",
)
args = parser.parse_args()
if args.level:
numeric_level = getattr(logging, args.level.upper(), None)
if not isinstance(numeric_level, int):
print(f"Invalid log level: {args.level}")
sys.exit(-1)
logging.basicConfig(level=numeric_level)
if numeric_level == logging.DEBUG:
osc.conf.config["debug"] = True
requests_log = logging.getLogger("requests.packages.urllib3")
requests_log.setLevel(logging.DEBUG)
requests_log.propagate = True
if not args.repodir:
args.repodir = pathlib.Path(args.package)
if args.repodir.exists() and not args.force:
print(f"Repository {args.repodir} already present")
sys.exit(-1)
elif args.repodir.exists() and args.force:
logging.info(f"Removing old repository {args.repodir}")
shutil.rmtree(args.repodir)
# TODO: use a CLI parameter to describe the projects
importer = Importer(
PROJECTS, args.package, args.repodir, args.search_ancestor, args.rebase_devel
)
importer.import_all_revisions(args.gc)
if __name__ == "__main__":
main()
Reference in New Issue View Git Blame Copy Permalink