Since pytest itself skips tests using dnspython if the latter is not
available, also using Automake conditionals for silently skipping
pytest-based tests requiring dnspython is redundant and hides
information. Allow all pytest-based tests requiring dnspython to be run
whenever pytest itself is available, in order to ensure test skipping is
done in a uniform manner.
Note that the above reasoning only applies to pytest-based tests, so
similar adjustments were not made for shell-based tests using Python
scripts that require dnspython ("chain", "cookie", "dnssec", "qmin").
The ability to conveniently mark tests which should only be run when the
CI_ENABLE_ALL_TESTS environment variable is set seems to be useful on a
general level and therefore it should not be limited to the "timeouts"
system test, where it is currently used.
pytest documentation [1] suggests to reuse commonly used test markers by
putting them all in a single Python module which then has to be imported
by test files that want to use the markers defined therein. Follow that
advice by creating a new bin/tests/system/pytest_custom_markers.py
Python module containing the relevant marker definitions.
Note that "import pytest_custom_markers" works from a test-specific
subdirectory because pytest modifies sys.path so that it contains the
paths to all parent directories containing a conftest.py file (and
bin/tests/system/ is one). PyLint does not like that, though, so add a
relevant PyLint suppression.
The above changes make bin/tests/system/timeouts/conftest.py redundant,
so remove it.
[1] https://docs.pytest.org/en/7.0.x/how-to/skipping.html#id1
Ensure all "import dns.*" statements are always placed after
pytest.importorskip('dns') calls, in order to allow the latter to
fulfill their purpose. Explicitly import all dnspython modules used by
each dnspython-based test to avoid relying on nested imports. Replace
function-scoped imports with global imports to reduce code duplication.
The intended purpose of the @pytest.mark.dnspython{,2} decorators was to
cause dnspython-based tests to be skipped if dnspython is not available
(or not recent enough). However, a number of system tests employing
those decorators contain global "import dns.resolver" statements which
trigger ImportError exceptions during test initialization if dnspython
is not available. In other words, the @pytest.mark.dnspython{,2}
decorators serve no useful purpose.
Currently, whenever a Python-based test requires dnspython, that
requirement applies to all tests in a given *.py file. Given that,
employ global pytest.importorskip() calls to ensure dnspython-based
parts of various system tests are skipped when dnspython is not
available. Remove all occurrences of the @pytest.mark.dnspython{,2}
decorators (and all associated code) to prevent confusion.
The intended purpose of the @pytest.mark.requests decorator was to cause
Python-based parts of the "statschannel" system test to be skipped if
the requests Python module is not available. However, both
tests-json.py and tests-xml.py contain a global "import requests"
statement which triggers ImportError exceptions during test
initialization if the requests module is not available. In other words,
the @pytest.mark.requests decorator serves no useful purpose.
Since all tests in both tests-json.py and tests-xml.py depend on the
requests Python module, employ pytest.importorskip() to ensure the
Python-based parts of the "statschannel" system test are skipped when
the requests module is not available. Remove all occurrences of the
@pytest.mark.requests decorator (and all associated code) to prevent
confusion.
All tests in bin/tests/system/statschannel/tests-xml.py require libxml2
support to be enabled in BIND 9 at build-time. Instead of applying the
same pytest.mark.skipif() decorator to every test in that file, set the
'pytestmark' global accordingly in order to immediately skip all tests
in tests-xml.py if libxml2 support is not compiled in.
Remove all occurrences of the @pytest.mark.xml decorator (and all
associated code) from the "statschannel" system test as the
xml.etree.ElementTree module is a part of the Python standard library
since Python 2.5 (so checking whether it is available is redundant) and
checking for libxml2 support in the tested BIND 9 build is already
handled by setting the 'pytestmark' global accordingly.
All tests in bin/tests/system/statschannel/tests-json.py require json-c
support to be enabled in BIND 9 at build-time. Instead of applying the
same pytest.mark.skipif() decorator to every test in that file, set the
'pytestmark' global accordingly in order to immediately skip all tests
in tests-json.py if json-c support is not compiled in.
Remove all occurrences of the @pytest.mark.json decorator (and all
associated code) from the "statschannel" system test as the json module
is a part of the Python standard library since Python 2.6 (so checking
whether it is available is redundant) and checking for json-c support in
the tested BIND 9 build is already handled by setting the 'pytestmark'
global accordingly.
Also remove a related excerpt from bin/tests/system/rpzextra/conftest.py
as it is a copy-paste artifact that serves no purpose in the "rpzextra"
system test.
The "statschannel" system test contains two Python helper modules:
- generic.py: test functions directly invoked by both tests-json.py
and test-xml.py,
- helper.py: helper functions invoked by test functions in generic.py.
The above logic for splitting helper functions into Python modules
prevents selective test skipping from working due to unconditional
import statements being present in both helper modules. For example, if
dnspython is not available on the test host, tests-json.py imports
generic.py, which in turn imports helper.py, which in turn attempts to
import various dnspython modules, triggering ImportError exceptions
during test initialization. Various decorators used for some tests
(like @pytest.mark.dnspython) suggest that such a scenario should be
handled gracefully, but that is not the case - modifying the test
collection in conftest.py does not prevent pytest from failing due to
import errors.
Fix by moving helper functions around to achieve a different split:
- generic.py: helper functions only relying on the Python standard
library,
- generic_dnspython.py: helper functions requiring dnspython.
Only two tests in tests-{json,xml}.py need dnspython to work
(test_traffic_json(), test_traffic_xml()). Since all
dnspython-dependent code is now present in generic_dnspython.py, employ
pytest.importorskip() in those two tests to ensure they can be
selectively skipped when dnspython is not available. Adjust other code
to account for the revised Python helper module layout. Remove all
occurrences of the @pytest.mark.dnspython decorator (and all associated
code) from the "statschannel" system test to prevent confusion.
The find invocation used by the bin/tests/system/get_ports.sh script
("find . -maxdepth 1 -mindepth 1 -type d") assumes the list of
directories in bin/tests/system/ remains unchanged throughout the run
time of a single system test suite. With pytest in use and the
conftest.py file now present in bin/tests/system/, that assumption is no
longer true as a __pycache__ directory may be created when the first
pytest-based test is started. Since the list of names returned by the
above find invocation serves as a fixed-size array of "port range
slots", any changes to that list during a system test suite run may lead
to port assignment collisions [1].
Fix by making the find invocation more nuanced, so that it only returns
names of directories containing test code. Squash a grep / cut pipeline
into a single awk invocation.
[1] see commit 31e5ca4bd9
Most Python-based system tests need to know which ports were assigned to
a given test by bin/tests/system/get_ports.sh. This is currently
handled by inspecting the values of various environment variables (set
by bin/tests/system/run.sh) and passing the port numbers to Python
scripts via pytest fixtures. However, this glue code has so far been
copy-pasted into each system test using it, rather than reused.
Since pytest also looks for conftest.py files in parent directories,
move commonly used fixtures to bin/tests/system/conftest.py. Set the
scope of all the moved fixtures to "session" as their return values are
only based on environment variables, so there is no point in recreating
them for every test requesting them. Adjust test code accordingly.
Previously, the function(s) in the commit subject could fail for various
reasons - mostly allocation failures, or other functions returning
different return code than ISC_R_SUCCESS. Now, the aforementioned
function(s) cannot ever fail and they would always return ISC_R_SUCCESS.
Change the function(s) to return void and remove the extra checks in
the code that uses them.
bad-ksk-without-zsk.conf only has a ksk defined without a
matching zsk for the same algorithm.
bad-zsk-without-ksk.conf only has a zsk defined without a
matching ksk for the same algorithm.
bad-unpaired-keys.conf has two keys of different algorithms
one ksk only and the other zsk only
The keep-response-order option has been introduced when TCP pipelining
has been introduced to BIND 9 as a failsafe for possibly non-compliant
clients.
Declare the keep-response-order obsolete as all DNS clients should
either support out-of-order processing or don't send more DNS queries
until the DNS response for the previous one has been received.
Extend the timeouts system test to ensure that the maximum outgoing
transfer time (max-transfer-time-out) and maximum outgoing transfer idle
time (max-transfer-idle-out) works as expected. This is done by
lowering the limits to 5/1 minutes and testing that the connection has
been dropped while sleeping between the individual XFR messages.
bin/tests/system/makejournal needs to ignore DNS_R_SEENINCLUDE
when calling dns_db_load(), otherwise it cannot generate a journal
for a zone file with a $INCLUDE statement.
Extend the timeouts system test that bursts the queries for large TXT
record and never read any responses back filling up the server TCP write
buffer. The test should work with the default wmem_max value on
Linux (208k).
There was a bug in the checking of the "blackhole" ACL in
dns_request_create*(), causing an address to be treated as included
in the ACL if it was explicitly *excluded*. Thus, leaving "blackhole"
unset had no effect, but setting it to "none" would cause any
destination addresses to be rejected for dns_request purposes. This
would cause zone transfer requests and SOA queries to fail, among
other things.
The bug has been fixed, and "blackhole { none; };" was added to the
xfer system test as a regression test.
The isc_thread_setaffinity call was removed in !5265 and we are not
going to restore it because it was proven that the performance is better
without it. Additionally, remove the already disabled cpu system test.
The isc_thread_setconcurrency function is unused and also calling
pthread_setconcurrency() on Linux has no meaning, formerly it was
added because of Solaris in 2001 and it was removed when taskmgr was
refactored to run on top of netmgr in !4918.
When there are more than one tokens initialized in SoftHSMv2,
care must be taken to correctly identify them.
Use a SoftHSMv2 token label which will uniquely identify the
token used for this test.
Use the "--token-label" parameter for the `pkcs11-tool` program
to make sure that it finds and uses the correct token.
The 'id' variable is either keyfromlabel-ksk or keyfromlabel-zsk and is
set in the 'keygen' and 'keyfromlabel' functions. It should not be used
outside these functions.
This test was originally in the pkcs11 system test. While this crash
happened in the native pkcs11 of BIND 9, and that code has been
removed in 9.17, there is no need for this test. Nevertheless, it
doesn't hurt having the test case persist.
Add a system test for engine_pkcs11 interactions that replaces the
tests that are done in the native PKCS#11 system test.
The native PKCS#11 code was removed in 9.17 but without copying the
pkcs11 system test.
Apparently we forgot about DLZ when updating DNS_CLIENTINFO_VERSION
constant for ECS, which is at value "3" since ECS was introduced.
The code in example drivers and tests now hardcodes version numbers
2 (without ECS) and 3 (with ECS) depending on what a given code path
requires.
System test used to have sequential system tests, which can't run in
parallel with the rest of system tests. As there are no such tests
anymore the underlying infrastructure can be dropped.
"parallel.sh" script was used on Windows to run system tests in
parallel. Since Windows support was removed from BIND 9, the script is
not needed anymore.
testsummary.sh was not updated after build system rewrite to Autotools,
and needs to be fixed to produce test summary and core dump, assertion
failures, and ThreadSanitizer reports.
Given that all of this is provided by Autotools and run.sh already,
there's little use to testsummary.sh script and should be dropped.
runall.sh was mainly used on Windows and as it's support was removed
from the "main" branch the script is not needed anymore.
Also, remove bin/tests/system/README text on running multiple system
test suites simultaneously with runall.sh as that support was not
present in the script anyway.
bin/tests/system/setup.sh just executes setup.sh script of a particular
system test in the directory of the system test. This does not seems to
be useful enough to maintain it.
stopall.sh script takes almost 2 minutes to go thru all test
subdirectories (due to a sleep in stop.pl) and does not seems to be
efficient way to stop manually started tests.
The keyfromlabel system ECDSA tests sometimes fail. When this happens
the ZSK and KSK key id values differ by 1, which is an indication that
the same key is used for both DNSKEY records.
When the private key is retrieved with 'ENGINE_load_private_key()', the
public key is already set. But sometimes that key differs from the key
which was retrieved with 'ENGINE_load_public_key()'.
The libp11 source code uses id to find the key and without IDs all the
keys are "equal", so it is returning the first key in the array of the
enumerated keys instead of the matching key. In our test we didn't use
'--id', just '--label'. With this change, the system test should no
longer fail intermittently.
Note this is only an issue for ECDSA keys, not RSA keys.
Add missing system test for dnssec-keyfromlabel. Test for various
algorithms that we can generate key files from a key that is stored in a
HSM, and that those keys can be used for signing with dnssec-signzone.
- Removed all code that only runs under CYGWIN, and made all
code that doesn't run under CYGWIN non-optional.
- Removed the $TP variable which was used to add optional
trailing dots to filenames; they're no longer optional.
- Removed references to pssuspend and dos2unix.
- No need to use environment variables for diff and kill.
- Removed uses of "tr -d '\r'"; this was a workaround for
a cygwin regex bug that is no longer needed.
Commit c787a539d2 fixed a certain class of
intermittent system test failures caused by named instances unable to
restart. The root cause was bin/tests/system/stop.pl returning without
waiting for a named instance to remove its lock file.
Later on, it turned out that the above change causes other issues on
Windows due to the way named handles signals on that platform. Commit
761ba4514f intended to address those
issues by making the server_lock_file() subroutine in
bin/tests/system/stop.pl return an empty value on Windows, in order to
prevent the script for waiting for lock file cleanup on that platform.
Note, however, that Windows detection in that subroutine is limited to
checking whether the CYGWIN environment variable is set.
While that environment variable was not set on Unix-like systems before
commit 761ba4514f, another commit
(a33237f070, merged a few weeks later)
changed that by setting the CYGWIN environment variable to an empty
value on Unix-like systems. This made the defined($ENV{'CYGWIN'}) check
in server_lock_file() return true, inadvertently preventing
bin/tests/system/stop.pl from waiting for lock file removal before
exiting on Unix-like systems and therefore reintroducing the original
issue.
Fix by making server_lock_file() only return an empty value when the
CYGWIN environment variable is set to a non-empty value (which is what
bin/tests/system/conf.sh.win32 does). Adjust a similar check in the
pid_file_exists() subroutine in the same way for consistency.
The echo_*() and cat_*() functions in bin/tests/system/conf.sh.common
call the "read" builtin command without specifying the field separator
to use. This results in leading whitespace getting stripped from each
line of the texts passed to those functions, which mangles e.g. pytest
output, hindering test failure troubleshooting.
Address by setting IFS to an empty value for the "read" calls used in
the aforementioned helper functions.
The bin/tests/system/start.pl script truncates the named.run file for a
given named instance unless it is invoked with the --restart
command-line option. Ever since Python-based tests were introduced,
bin/tests/system/run.sh may start named instances used by a given system
test multiple times within a single run, causing the
bin/tests/system/start.pl script to truncate some of the log files
written during the test. This makes troubleshooting certain test
failures hard or even impossible.
Fix by calling bin/tests/system/start.pl with the --restart command-line
option for every start_servers() invocation except the first one.
When failure is expected, the `rndc` command in the catz system test
is being called directly instead of using a function, i.e.:
$RNDC -c ../common/rndc.conf -s 10.53.0.2 -p 9953 reconfig \
> /dev/null 2>&1 && ret=1
... instead of:
rndccmd 10.53.0.2 reconfig && ret=1
This is done to suppress messages like "lt-rndc: 'reconfig' failed:
failure" appearing in the message log of the test, because failure
is actually expected, and the appearance of that message can be
confusing.
The port value used in this case is not correct, making the
`rndc reload` command to fail. This error was not detected earlier
only because the failure of the command is actually expected, but
the failure happens for a "wrong" reason, and the test still passes.
Fix the error by using the existing variable instead of the fixed
number.
