Differential D52

implement yumrepoinfo
ClosedPublic
Actions

Authored by kparal on Apr 14 2014, 2:14 PM.

Details

Reviewers
tflink
jskladan
Maniphest Tasks
T111: Port repoinfo.conf from AutoQA to Taskotron
Commits
rLTRN34e39154cccb: implement yumrepoinfo
Summary

This ports repoinfo.conf from AutoQA into Taskotron as yumrepoinfo. A lot of adjustments were made. Most important notes:

  1. It again uses the singleton pattern (using a global variable) to avoid parsing config files for every instantiation.
  2. I decided to keep the ConfigParser structure because the string interpolation is convenient. But if we think we should e.g. convert everything to YAML, it can be done.
Test Plan

New tests. Unlike with config.py, I decided to allow config file parsing from disk even for unit tests. The rationale:

  • with Config, it doesn't make sense to load adjustments for test suite (the test results should be the same everywhere => reliable)
  • with Config, we already have a set of defaults in config_defaults.py (and we need them), so we can easily just load that
  • with yumrepoinfo, I'm unclear whether some unit tests (for different modules) won't be written with a particular repoinfo contents in mind. Anything statically provided could quickly become obsolete (yumrepoinfo changes at least once every 6 months)
  • with yumrepoinfo, in order to provide the defaults without config file parsing, I'd have to create yumrepoinfo_defaults.py, which would need to be imported and parsed, therefore causing additional disk access anyway.

So, I think there's no performance hit when allowing other test modules to import yumrepoinfo and load the config file from disk. Because of singleton pattern, it's done only once (per arch). For some heavier disk activity, functional tests should be used, of course.

Diff Detail

Lint
Lint Skipped
Unit
Unit Tests Skipped
jskladan requested changes to this revision.Apr 15 2014, 12:39 PM
jskladan added inline comments.
libtaskotron/yumrepoinfo.py
51

Would it make sense to list the possible keys, that can be used?

54

I know this is a corner case, but maybe something like:

self.defaults['arch'] = rpm_utils.basearch(self.defaults.get('arch', arch))

would make sense. But listing the possible keys in defaults (see line 50), can easily solve the problem.

58

Not sure what the conventions are here, but I consider:

self.parser = ConfigParser.SafeConfigParser(defaults = self.defaults)

more readable.

kparal added inline comments.Apr 15 2014, 1:06 PM
libtaskotron/yumrepoinfo.py
51

I have decided to remove the option completely, nobody has ever used it in AutoQA. We can add it back once needed. Thanks for pointing out the problems.

58

Sorry to disagree :-)
http://legacy.python.org/dev/peps/pep-0008/#other-recommendations

We don't need to follow PEP8 guidelines, not at all, but in this case I actually consider missing spaces more readable than present spaces, in line with the recommendation. So that's why I write it this way.

kparal updated this revision.Apr 15 2014, 1:10 PM
  • remove defaults= from init, because nobody has ever used it
  • add explanation header to yumrepoinfo.conf
kparal updated this revision.Apr 15 2014, 1:12 PM
  • remove init(defaults=) unit test
tflink accepted this revision.Apr 16 2014, 7:38 AM

Sigh, another piece of code that I don't really like but also don't have a better solution for. I don't like the idea of statically defining valid releases in a config file that we're going to have to package. That leaves us in a similar situation as AutoQA where the package will need to be updated every time the repos change but on the other hand, there are other packages that are also required to update as releases change state (mock is one that comes to mind). At least this way, we aren't mandating the existance of /etc/taskotron/yumrepoinfo.conf which was a frustration of mine with AutoQA's repoinfo.

with Config, it doesn't make sense to load adjustments for test suite (the test results should be the same everywhere => reliable)

Sure, but you still run into the potential issue where production will almost always be using a different absolute repo than anything outside the fedora infra. It's also using download.fp.o which returns a different mirror for almost every call - I'm not sure that we're really ensuring consistency here.

with yumrepoinfo, I'm unclear whether some unit tests (for different modules) won't be written with a particular repoinfo contents in mind. Anything statically provided could quickly become obsolete (yumrepoinfo changes at least once every 6 months)

Why are we statically defining all of this? All of this _should_ be algorithmic in the general case so we should be able to have some templates for yum repos that are fed a baseurl and release (special casing rawhide) to get our yum repo information instead of having everything live in a config file.

with yumrepoinfo, in order to provide the defaults without config file parsing, I'd have to create yumrepoinfo_defaults.py, which would need to be imported and parsed, therefore causing additional disk access anyway.

This sounds like premature optimization to me. How many times would repoinfo be parsed per task execution? I can't imagine any more than 2 or 3 unless we end up with some poorly written tasks. The singleton can't cross exec() boundaries unless explicitly passed so unless it's only being called inside tasks, it'll be parsed multiple times either way.

So, in the end, I'm not crazy about this but I don't have a better idea. With that in mind, it looks good to me. Thanks for taking care of this.

conf/yumrepoinfo.conf
9

I'd like to see a comment here about replacing download.fedoraproject.org. Outside of the fedora infra, it is a frontend to mirrormanager and will not always return the same mirror for succesive requests. I think it's the best default we could have but also think that adding a note that it would be better to replace it with a semi-local absolute mirror would be best

kparal added a comment.Apr 16 2014, 12:55 PM

Sigh, another piece of code that I don't really like but also don't have a better solution for. I don't like the idea of statically defining valid releases in a config file that we're going to have to package. That leaves us in a similar situation as AutoQA where the package will need to be updated every time the repos change but on the other hand, there are other packages that are also required to update as releases change state (mock is one that comes to mind).

It's very suboptimal, yes. I tried to port an existing solution, not to devise a new one. But if you have some good pointers, I can try. When I think about it, I could have made most of the document dynamic. The only static pieces that need to be updated every 6 months are:

  1. Rawhide to tag mapping:
[rawhide]
tag = fc21
  1. Stable vs development status for each repo. This status changes a) URLs b) probably also some additional semantics some checks might need to have.

I don't really know how to avoid the need to have a configuration file defining at least these two pieces of information and update it regularly.

All the other stuff can probably go away and I can do it if you want. The logic will be in code instead of in a config file, but the logic is very stable (how you compute URL for a repo, or which repo is a parent of which repo). As long as we support just Fedora.

with Config, it doesn't make sense to load adjustments for test suite (the test results should be the same everywhere => reliable)

Sure, but you still run into the potential issue where production will almost always be using a different absolute repo than anything outside the fedora infra. It's also using download.fp.o which returns a different mirror for almost every call - I'm not sure that we're really ensuring consistency here.

With the current setup, any config file modifications will have no effect on the test suite. Are you trying to say they should? For unit tests, I think the URLs don't matter (we won't download anything anyway). For functest, there might be such a need... not sure. I assumed it's better to deal with it once we get there.

Why are we statically defining all of this? All of this _should_ be algorithmic in the general case so we should be able to have some templates for yum repos that are fed a baseurl and release (special casing rawhide) to get our yum repo information instead of having everything live in a config file.

Yes, if we put the logic into code (we take a URL template from the config file and substitute a few variables in it, that's it). We still need to know which release is stable and which is development, though, as mentioned above. This is probably a simple change to do, if you prefer it that way, I can change it.

conf/yumrepoinfo.conf
9

I don't think we had many problems with it, because as long as yum operates on those URLs, it resolves the server first and then performs all operations with a single repo (see baseurl= in /etc/yum.repos.d/fedora.repo, it's the same). But yeah, I remember some kind of issues like that in the past. I'll add a note that it's better to hand-pick a mirror.

kparal closed this revision.Apr 16 2014, 1:18 PM

Closed by commit rLTRN34e39154cccb (authored by @kparal).

Revision Contents

Diff 158

View Options

conf/yumrepoinfo.conf

  • This file was added.
1​## A Taskotron configuration file representing available YUM repositories inside
2​## Fedora infrastructure.
3​## This file is in a ConfigParser syntax, very similar to INI syntax known from
4​## Windows.
5​
6​[DEFAULT]
7​# URLs to yum repos
8​# Please note: 'download.fedoraproject.org' is a redirect that points you to a
9​# random mirror every time. For faster and more reliable operation, pick a close
tflinkUnsubmitted
Not Done

I'd like to see a comment here about replacing download.fedoraproject.org. Outside of the fedora infra, it is a frontend to mirrormanager and will not always return the same mirror for succesive requests. I think it's the best default we could have but also think that adding a note that it would be better to replace it with a semi-local absolute mirror would be best

kparalAuthorUnsubmitted
Not Done

I don't think we had many problems with it, because as long as yum operates on those URLs, it resolves the server first and then performs all operations with a single repo (see baseurl= in /etc/yum.repos.d/fedora.repo, it's the same). But yeah, I remember some kind of issues like that in the past. I'll add a note that it's better to hand-pick a mirror.

10​# mirror from http://mirrors.fedoraproject.org and use it in 'baseurl'.
11​baseurl = http://download.fedoraproject.org/pub/fedora/linux
12​goldurl = %(baseurl)s/releases/%(path)s/Everything/%(arch)s/os
13​updatesurl = %(baseurl)s/updates/%(path)s/%(arch)s
14​rawhideurl = %(baseurl)s/%(path)s/%(arch)s/os
15​
16​# list of supported architectures
17​arches = i386, x86_64
18​
19​# a direct parent of this repo, essential for operations (dependencies)
20​parent =
21​
22​# koji tag defaults to section name
23​tag = %(__name__)s
24​
25​# true for "top" repos corresponding to currently supported Fedora releases
26​supported = no
27​
28​
29​# Rawhide
30​[rawhide]
31​path = development/rawhide
32​url = %(rawhideurl)s
33​tag = f21
34​
35​# Fedora 20
36​[f20]
37​url = %(goldurl)s
38​path = 20
39​supported = yes
40​
41​[f20-updates]
42​url = %(updatesurl)s
43​path = 20
44​parent = f20
45​
46​[f20-updates-testing]
47​url = %(updatesurl)s
48​path = testing/20
49​parent = f20-updates
50​
51​# Fedora 19
52​[f19]
53​url = %(goldurl)s
54​path = 19
55​supported = yes
56​
57​[f19-updates]
58​url = %(updatesurl)s
59​path = 19
60​parent = f19
61​
62​[f19-updates-testing]
63​url = %(updatesurl)s
64​path = testing/19
65​parent = f19-updates
66​
View Options

libtaskotron/yumrepoinfo.py

  • This file was added.
1​# -*- coding: utf-8 -*-
2​# Copyright 2009-2014, Red Hat, Inc.
3​# License: GNU General Public License version 2 or later
4​
5​'''A wrapper object for yumrepoinfo.conf to access its information easily'''
6​
7​import ConfigParser
8​import os
9​
10​from . import config
11​from .logger import log
12​from . import rpm_utils
13​from . import exceptions as exc
14​
15​
16​# a singleton instance of YumRepoInfo (a dict with arches as keys)
17​_yumrepoinfo = {}
18​
19​
20​def get_yumrepoinfo(arch=None):
21​ '''Get YumRepoInfo instance.
22​
23​ @param arch architecture to return the YumRepoInfo for. It's always
24​ converted to basearch. If None, then local machine arch is used.
25​ @return shared YumRepoInfo instance. This method is implemented using the
26​ singleton pattern - you will always receive the same instance, which
27​ will get auto-initialized on the first method call.
28​ @throw TaskotronConfigError if file config parsing and handling failed
29​ '''
30​ # converts to basearch, returns local arch for None
31​ arch = rpm_utils.basearch(arch)
32​
33​ if not arch in _yumrepoinfo:
34​ _yumrepoinfo[arch] = YumRepoInfo(arch)
35​ return _yumrepoinfo[arch]
36​
37​
38​class YumRepoInfo(object):
39​ '''This class is a wrapper for easily accessing repoinfo.conf file.'''
40​
41​ def __init__(self, arch=None, filelist=None):
42​ '''Create YumRepoinfoConfig object.
43​
44​ @param arch architecture for which to adjust repo URLs. By default it
45​ refers to the architecture of the current machine. It's
46​ always converted to basearch.
47​ @param filelist list of config files to read information from. The first
48​ available config file is used. If None, then the default
49​ list of locations is used.
50​ '''
51​ self.arch = rpm_utils.basearch(arch)
jskladanUnsubmitted
Not Done

Would it make sense to list the possible keys, that can be used?

kparalAuthorUnsubmitted
Not Done

I have decided to remove the option completely, nobody has ever used it in AutoQA. We can add it back once needed. Thanks for pointing out the problems.

52​ self.filelist = (filelist if filelist is not None else
53​ [os.path.join(confdir, 'yumrepoinfo.conf')
54​ for confdir in config.CONF_DIRS])
jskladanUnsubmitted
Not Done

I know this is a corner case, but maybe something like:

self.defaults['arch'] = rpm_utils.basearch(self.defaults.get('arch', arch))

would make sense. But listing the possible keys in defaults (see line 50), can easily solve the problem.

55​ self.parser = ConfigParser.SafeConfigParser(defaults=
56​ {'arch': self.arch})
57​
58​ self._read()
jskladanUnsubmitted
Not Done

Not sure what the conventions are here, but I consider:

self.parser = ConfigParser.SafeConfigParser(defaults = self.defaults)

more readable.

kparalAuthorUnsubmitted
Not Done

Sorry to disagree :-)
http://legacy.python.org/dev/peps/pep-0008/#other-recommendations

We don't need to follow PEP8 guidelines, not at all, but in this case I actually consider missing spaces more readable than present spaces, in line with the recommendation. So that's why I write it this way.

59​
60​ if not self.repos():
61​ log.warn("No repos found in yumrepoinfo config files: %s",
62​ self.filelist)
63​
64​
65​ def repos(self):
66​ '''Get the list of all known repository names.'''
67​ return self.parser.sections()
68​
69​
70​ def releases(self):
71​ '''Get the list of stable (supported) Fedora releases.'''
72​ return [r for r in self.repos() if self.parser.getboolean(r,
73​ 'supported')]
74​
75​
76​ def arches(self, reponame):
77​ '''Get the list of all supported architectures for a repo'''
78​ return self._getlist(reponame, 'arches')
79​
80​
81​ def get(self, reponame, key):
82​ '''Get a specific key value from a repo
83​
84​ @throw TaskotronConfigError if the key can't be retrieved (e.g. wrong
85​ key or repo name)
86​ '''
87​ try:
88​ return self.parser.get(reponame, key)
89​ except ConfigParser.Error as e:
90​ raise exc.TaskotronConfigError("Can't retrieve key '%s' from repo "
91​ "'%s': %s" % (key, reponame, e))
92​
93​
94​ def repo(self, reponame):
95​ '''Given a repo name, return the yumrepoinfo dict, with keys:
96​ 'arches', 'parents', 'tag', 'url', 'path' and 'name' '''
97​ repo = {'arches': self.arches(reponame),
98​ 'parent': self.get(reponame, 'parent'),
99​ 'tag': self.get(reponame,'tag'),
100​ 'url': self.get(reponame,'url'),
101​ 'path': self.get(reponame,'path'),
102​ 'name': reponame,
103​ }
104​ return repo
105​
106​
107​ def repo_by_tag(self, tag):
108​ '''Given a Koji tag, return the corresponding repo dict.
109​
110​ @param tag a koji tag, e.g. 'f20-updates'. Note: 'rawhide' is not used
111​ as a Koji tag, use number identifier instead or use
112​ repo('rawhide')['tag'] to discover it first.
113​ @return repo dict as from @method repo(), or None if no such Koji tag
114​ found
115​ '''
116​ # we don't have repo definition for -candidate tags,
117​ # let's go up in the hierarchy (stripping '-candidate')
118​ if tag.endswith('-candidate'):
119​ tag = tag[:-len('-candidate')]
120​ for r in self.repos():
121​ if self.get(r,'tag') == tag:
122​ return self.repo(r)
123​
124​
125​ def top_parent(self, reponame):
126​ '''Go through the repo hiearchy and find the top parent for a repo
127​
128​ @param reponame repository name
129​ @return the top parent reponame, or None if there is no parent at all
130​ @throw TaskotronConfigError if infinite parent loop detected
131​ '''
132​ parents = []
133​ repo = reponame
134​
135​ while True:
136​ parent = self.get(repo, 'parent')
137​
138​ if not parent:
139​ if repo is reponame:
140​ # there is no parent at all, return None
141​ return None
142​ else:
143​ # we've found at least one parent
144​ return repo
145​
146​ # detect infinite parent loops
147​ if parent in parents:
148​ raise exc.TaskotronConfigError('Infinite parent loop detected '
149​ 'in yumrepoinfo: %s' % parents)
150​
151​ parents.append(parent)
152​ repo = parent
153​
154​
155​ def _read(self):
156​ '''Read first available config file from the list of provided config
157​ files.'''
158​ log.debug('Looking for yumrepoinfo config files in: %s', self.filelist)
159​
160​ for cfg in self.filelist:
161​ if self.parser.read(cfg):
162​ log.debug('Successfully loaded yumrepoinfo config file: %s',
163​ cfg)
164​ break
165​ else:
166​ log.debug('Failed to load yumrepoinfo config file: %s', cfg)
167​
168​
169​ def _getlist(self, reponame, item):
170​ '''Transform a list of comma separated values into a Python list
171​ Returns empty list for non-existent item'''
172​ itemlist = self.get(reponame, item)
173​ if not itemlist:
174​ return []
175​ else:
176​ return [t.strip() for t in itemlist.split(',')]
View Options

testing/functest_yumrepoinfo.py

  • This file was added.
1​# -*- coding: utf-8 -*-
2​# Copyright 2014, Red Hat, Inc.
3​# License: GNU General Public License version 2 or later
4​
5​'''Functional tests for libtaskotron/yumrepoinfo.py'''
6​
7​import pytest
8​
9​from libtaskotron import yumrepoinfo
10​from libtaskotron import exceptions as exc
11​
12​class TestYumRepoInfo(object):
13​
14​ def test_init_filelist(self, tmpdir):
15​ '''test YumRepoInfo(filelist=) with custom config files'''
16​ conf1 = tmpdir.join('conf1.cfg')
17​ conf2 = tmpdir.join('conf2.cfg')
18​ conf1.write('''
19​[DEFAULT]
20​a = foo''')
21​ conf2.write('''
22​[DEFAULT]
23​b = bar''')
24​ repoinfo = yumrepoinfo.YumRepoInfo(filelist=[conf1.strpath,
25​ conf2.strpath])
26​
27​ # must read conf1
28​ assert repoinfo.get('DEFAULT', 'a') == 'foo'
29​
30​ # must not read conf2
31​ with pytest.raises(exc.TaskotronConfigError):
32​ repoinfo.get('DEFAULT', 'b')
33​
34​ # must not read default configs
35​ with pytest.raises(exc.TaskotronConfigError):
36​ repoinfo.get('DEFAULT', 'tag')
View Options

testing/test_yumrepoinfo.py

  • This file was added.
1​# -*- coding: utf-8 -*-
2​# Copyright 2014, Red Hat, Inc.
3​# License: GNU General Public License version 2 or later
4​
5​'''Unit tests for libtaskotron/yumrepoinfo.py'''
6​
7​import pytest
8​import StringIO
9​import ConfigParser
10​
11​from libtaskotron import yumrepoinfo
12​from libtaskotron import exceptions as exc
13​
14​
15​TEST_CONF='''\
16​[DEFAULT]
17​baseurl = http://download.fedoraproject.org/pub/fedora/linux
18​goldurl = %(baseurl)s/releases/%(path)s/Everything/%(arch)s/os
19​updatesurl = %(baseurl)s/updates/%(path)s/%(arch)s
20​rawhideurl = %(baseurl)s/%(path)s/%(arch)s/os
21​arches = i386, x86_64
22​parent =
23​tag = %(__name__)s
24​supported = no
25​
26​[rawhide]
27​path = development/rawhide
28​url = %(rawhideurl)s
29​tag = f21
30​
31​[f20]
32​url = %(goldurl)s
33​path = 20
34​supported = yes
35​
36​[f20-updates]
37​url = %(updatesurl)s
38​path = 20
39​parent = f20
40​arches = x86_64
41​
42​[f20-updates-testing]
43​url = %(updatesurl)s
44​path = testing/20
45​parent = f20-updates
46​'''
47​
48​
49​class TestYumRepoInfo(object):
50​
51​ @classmethod
52​ def setup_class(cls):
53​ '''One-time class initialization'''
54​ # create YumRepoInfo initialized with TEST_CONF
55​ cls.repoinfo = yumrepoinfo.YumRepoInfo(filelist=[])
56​ cls.repoinfo.parser.readfp(StringIO.StringIO(TEST_CONF))
57​
58​ @pytest.fixture
59​ def disable_disk_loading(self, monkeypatch):
60​ '''Patch ConfigParser not to load from disk.'''
61​
62​ # instead of reading from disk, just return the input parameter
63​ monkeypatch.setattr(ConfigParser.RawConfigParser, 'read',
64​ lambda self, cfg: cfg)
65​
66​ @pytest.fixture
67​ def clean_singleton(self, monkeypatch):
68​ '''Replace singleton container with an empty dict'''
69​ monkeypatch.setattr(yumrepoinfo, '_yumrepoinfo', {})
70​
71​ def test_init_arch(self, disable_disk_loading):
72​ '''test YumRepoInfo(arch=)'''
73​ repoinfo = yumrepoinfo.YumRepoInfo(arch='x86_64')
74​ repoinfo.parser.readfp(StringIO.StringIO(TEST_CONF))
75​ assert '/x86_64/' in repoinfo.get('rawhide', 'goldurl')
76​
77​ # i686 should be converted to i386
78​ repoinfo = yumrepoinfo.YumRepoInfo(arch='i686')
79​ repoinfo.parser.readfp(StringIO.StringIO(TEST_CONF))
80​ assert '/i386/' in repoinfo.get('rawhide', 'goldurl')
81​
82​ def test_init_empty_filelist(self, monkeypatch):
83​ '''YumRepoInfo(filelist=[]) should not load anything from disk'''
84​ # make sure the disk is not read
85​ def _raise_alarm(x):
86​ assert False, 'This must not be called'
87​ monkeypatch.setattr(ConfigParser.RawConfigParser, 'read', _raise_alarm)
88​
89​ repoinfo = yumrepoinfo.YumRepoInfo(filelist=[])
90​ assert not repoinfo.repos()
91​
92​ def test_repos(self):
93​ '''repos() should return all known repo names'''
94​ assert self.repoinfo.repos() == ['rawhide', 'f20', 'f20-updates',
95​ 'f20-updates-testing']
96​
97​ def test_releases(self):
98​ '''releases() should return only stable releases repo names'''
99​ assert self.repoinfo.releases() == ['f20']
100​
101​ def test_arches(self):
102​ '''arches() should return what's defined in a section'''
103​ assert self.repoinfo.arches('f20') == ['i386', 'x86_64']
104​ assert self.repoinfo.arches('DEFAULT') == ['i386', 'x86_64']
105​ assert self.repoinfo.arches('f20-updates') == ['x86_64']
106​
107​ def test_get(self):
108​ '''get() simply returns a key'''
109​ assert self.repoinfo.get('rawhide', 'supported') == 'no'
110​ assert self.repoinfo.get('DEFAULT', 'parent') == ''
111​
112​ def test_get_raise(self):
113​ '''get() should raise an error when the key doesn't exist'''
114​ with pytest.raises(exc.TaskotronConfigError):
115​ self.repoinfo.get('rawhide', 'non-existent key')
116​
117​ with pytest.raises(exc.TaskotronConfigError):
118​ self.repoinfo.get('non-existent repo', 'tag')
119​
120​ def test_repo(self):
121​ '''repo() returns a repo dict'''
122​ rawhide = self.repoinfo.repo('rawhide')
123​ assert rawhide['name'] == 'rawhide'
124​ assert rawhide['tag'] == 'f21'
125​ assert set(rawhide.keys()).issuperset(
126​ set(('arches', 'parent', 'tag', 'url', 'path', 'name'))
127​ )
128​
129​ def test_repo_by_tag(self):
130​ '''repo_by_tag() must work based on tags, not section names'''
131​ f20up = self.repoinfo.repo_by_tag('f20-updates')
132​ assert f20up['name'] == f20up['tag'] == 'f20-updates'
133​
134​ f21 = self.repoinfo.repo_by_tag('f21')
135​ assert f21['tag'] == 'f21'
136​ assert f21['name'] == 'rawhide'
137​
138​ assert self.repoinfo.repo_by_tag('foobar') is None
139​
140​ def test_top_parent(self):
141​ '''top_parent() must travel the hierarchy'''
142​ assert self.repoinfo.top_parent('rawhide') is None
143​ assert self.repoinfo.top_parent('f20-updates') == 'f20'
144​ assert self.repoinfo.top_parent('f20-updates-testing') == 'f20'
145​
146​ def test_top_parent_raise(self):
147​ '''top_parent() must detect an infinite parent cycle'''
148​ cfg = '''\
149​[repo1]
150​parent = repo2
151​[repo2]
152​parent = repo1
153​'''
154​ repoinfo = yumrepoinfo.YumRepoInfo(filelist=[])
155​ repoinfo.parser.readfp(StringIO.StringIO(cfg))
156​
157​ with pytest.raises(exc.TaskotronConfigError):
158​ repoinfo.top_parent('repo1')
159​
160​ def test_get_yumrepoinfo(self, clean_singleton):
161​ '''get_yumrepoinfo must work as a singleton'''
162​
163​ repoinfo = yumrepoinfo.get_yumrepoinfo()
164​ # two calls with same arch must return the same instance
165​ assert repoinfo is yumrepoinfo.get_yumrepoinfo()
166​ # there should be a record in the singleton container
167​ assert len(yumrepoinfo._yumrepoinfo.keys()) > 0
168​
169​ # the same must work for basearch conversion
170​ repoinfo_i686 = yumrepoinfo.get_yumrepoinfo(arch='i686')
171​ repoinfo_i386 = yumrepoinfo.get_yumrepoinfo(arch='i386')
172​ assert repoinfo_i686 is repoinfo_i386
173​ assert 'i386' in yumrepoinfo._yumrepoinfo
174​ assert 'i686' not in yumrepoinfo._yumrepoinfo