X-Git-Url: https://git.arvados.org/arvados-dev.git/blobdiff_plain/fcc0a767df83518e1b42c78ab7d4805fe6b5192b..refs/heads/19092-upload-crunchstat_summary-to-pypi:/jenkins/run_upload_packages.py diff --git a/jenkins/run_upload_packages.py b/jenkins/run_upload_packages.py index eff30cc..6f695b7 100755 --- a/jenkins/run_upload_packages.py +++ b/jenkins/run_upload_packages.py @@ -1,19 +1,65 @@ #!/usr/bin/env python3 +# Copyright (C) The Arvados Authors. All rights reserved. +# +# SPDX-License-Identifier: AGPL-3.0 + import argparse +import errno import functools import glob +import locale import logging import os import pipes +import re import shutil import subprocess import sys import time +def run_and_grep(cmd, read_output, *regexps, + encoding=locale.getpreferredencoding(), **popen_kwargs): + """Run a subprocess and capture output lines matching regexps. + + Arguments: + * cmd: The command to run, as a list or string, as for subprocess.Popen. + * read_output: 'stdout' or 'stderr', the name of the output stream to read. + Remaining arguments are regexps to match output, as strings or compiled + regexp objects. Output lines matching any regexp will be captured. + + Keyword arguments: + * encoding: The encoding used to decode the subprocess output. + Remaining keyword arguments are passed directly to subprocess.Popen. + + Returns 2-tuple (subprocess returncode, list of matched output lines). + """ + regexps = [regexp if hasattr(regexp, 'search') else re.compile(regexp) + for regexp in regexps] + popen_kwargs[read_output] = subprocess.PIPE + proc = subprocess.Popen(cmd, **popen_kwargs) + with open(getattr(proc, read_output).fileno(), encoding=encoding) as output: + matched_lines = [] + for line in output: + if any(regexp.search(line) for regexp in regexps): + matched_lines.append(line) + if read_output == 'stderr': + print(line, file=sys.stderr, end='') + return proc.wait(), matched_lines + + class TimestampFile: def __init__(self, path): self.path = path + # Make sure the dirname for `path` exists + p = os.path.dirname(path) + try: + os.makedirs(p) + except OSError as exc: + if exc.errno == errno.EEXIST and os.path.isdir(p): + pass + else: + raise self.start_time = time.time() def last_upload(self): @@ -23,9 +69,18 @@ class TimestampFile: return -1 def update(self): - os.close(os.open(self.path, os.O_CREAT | os.O_APPEND)) - os.utime(self.path, (time.time(), self.start_time)) - + try: + os.close(os.open(self.path, os.O_CREAT | os.O_APPEND)) + os.utime(self.path, (time.time(), self.start_time)) + except: + # when the packages directory is created/populated by a build in a + # docker container, as root, the script that runs the upload + # doesn't always have permission to touch a timestamp file there. + # In production, we build/upload from ephemeral machines, which + # means that the timestamp mechanism is not used. We print a + # warning and move on without erroring out. + print("Warning: unable to update timestamp file",self.path,"permission problem?") + pass class PackageSuite: NEED_SSH = False @@ -62,6 +117,14 @@ class PackageSuite: class PythonPackageSuite(PackageSuite): LOGGER_PART = 'python' + REUPLOAD_REGEXPS = [ + re.compile( + r'^error: Upload failed \(400\): A file named "[^"]+" already exists\b'), + re.compile( + r'^error: Upload failed \(400\): File already exists\b'), + re.compile( + r'^error: Upload failed \(400\): Only one sdist may be uploaded per release\b'), + ] def __init__(self, glob_root, rel_globs): super().__init__(glob_root, rel_globs) @@ -72,37 +135,33 @@ class PythonPackageSuite(PackageSuite): if src_dir in self.seen_packages: return self.seen_packages.add(src_dir) - # NOTE: If we ever start uploading Python 3 packages, we'll need to - # figure out some way to adapt cmd to match. It might be easiest - # to give all our setup.py files the executable bit, and run that - # directly. # We also must run `sdist` before `upload`: `upload` uploads any # distributions previously generated in the command. It doesn't # know how to upload distributions already on disk. We write the # result to a dedicated directory to avoid interfering with our # timestamp tracking. - cmd = ['python2.7', 'setup.py'] + cmd = ['python3', 'setup.py'] if not self.logger.isEnabledFor(logging.INFO): cmd.append('--quiet') - cmd.extend(['sdist', '--dist-dir', '.upload_dist', 'upload']) - subprocess.check_call(cmd, cwd=src_dir) + cmd.extend(['bdist_wheel', '--dist-dir', '.upload_dist']) + cmd.extend(['sdist', '--dist-dir', '.upload_dist']) + cmd.extend(['upload']) + upload_returncode, repushed = run_and_grep( + cmd, 'stderr', *self.REUPLOAD_REGEXPS, cwd=src_dir) + if (upload_returncode != 0) and not repushed: + raise subprocess.CalledProcessError(upload_returncode, cmd) shutil.rmtree(os.path.join(src_dir, '.upload_dist')) class GemPackageSuite(PackageSuite): LOGGER_PART = 'gems' + REUPLOAD_REGEXP = re.compile(r'^Repushing of gem versions is not allowed\.$') def upload_file(self, path): cmd = ['gem', 'push', path] - push_proc = subprocess.Popen(cmd, stdout=subprocess.PIPE) - repushed = any(line == b'Repushing of gem versions is not allowed.\n' - for line in push_proc.stdout) - # Read any remaining stdout before closing. - for line in push_proc.stdout: - pass - push_proc.stdout.close() - if (push_proc.wait() != 0) and not repushed: - raise subprocess.CalledProcessError(push_proc.returncode, cmd) + push_returncode, repushed = run_and_grep(cmd, 'stdout', self.REUPLOAD_REGEXP) + if (push_returncode != 0) and not repushed: + raise subprocess.CalledProcessError(push_returncode, cmd) class DistroPackageSuite(PackageSuite): @@ -136,28 +195,44 @@ class DistroPackageSuite(PackageSuite): self.__class__.__name__, *(pipes.quote(s) for s in args))) def upload_files(self, paths): + dest_dir = os.path.join(self.REMOTE_DEST_DIR, self.target) + mkdir = self._build_cmd('ssh', self.ssh_host, 'install', '-d', dest_dir) + subprocess.check_call(mkdir) cmd = self._build_cmd('scp', *paths) - cmd.append('{self.ssh_host}:{self.REMOTE_DEST_DIR}'.format(self=self)) + cmd.append('{}:{}'.format(self.ssh_host, dest_dir)) subprocess.check_call(cmd) class DebianPackageSuite(DistroPackageSuite): - FREIGHT_SCRIPT = """ + APT_SCRIPT = """ cd "$1"; shift DISTNAME=$1; shift -freight add "$@" "apt/$DISTNAME" -freight cache "apt/$DISTNAME" -rm "$@" +for package in "$@"; do + set +e + aptly repo search "$DISTNAME" "${package%.deb}" >/dev/null 2>&1 + RET=$? + set -e + if [[ $RET -eq 0 ]]; then + echo "Not adding $package, it is already present in repo $DISTNAME" + rm "$package" + else + aptly repo add -remove-files "$DISTNAME" "$package" + fi +done +aptly publish update "$DISTNAME" filesystem:"${DISTNAME%-*}": """ - TARGET_DISTNAMES = { - 'debian7': 'wheezy', - 'debian8': 'jessie', - 'ubuntu1204': 'precise', - 'ubuntu1404': 'trusty', - } + + def __init__(self, glob_root, rel_globs, target, ssh_host, ssh_opts, repo): + super().__init__(glob_root, rel_globs, target, ssh_host, ssh_opts) + self.TARGET_DISTNAMES = { + 'debian10': 'buster-'+repo, + 'debian11': 'bullseye-'+repo, + 'ubuntu1804': 'bionic-'+repo, + 'ubuntu2004': 'focal-'+repo, + } def post_uploads(self, paths): - self._run_script(self.FREIGHT_SCRIPT, self.REMOTE_DEST_DIR, + self._run_script(self.APT_SCRIPT, self.REMOTE_DEST_DIR + '/' + self.target, self.TARGET_DISTNAMES[self.target], *self._paths_basenames(paths)) @@ -168,17 +243,20 @@ cd "$1"; shift REPODIR=$1; shift rpmsign --addsign "$@"