The previous symlink name convention was prone to name collisions If a
system test contained both a shell test and a pytest module of the same
name (e.g. dnstap test has both tests.sh and tests_dnstap.py), then
these would have the same convenience symlink, which could cause test
setup issues as well as confusion when examining test artifacts.
Update the naming convention to include the full pytest module name.
This results in a slightly more verbose names for shell tests (e.g.
dnstap_sh_dnstap instead of the previous dnstap_dnstap), but it removes
the chance of a collision.
Reorganize individual port fixtures and re-use the ports fixture to
obtain their number. Store it as integer and only cast it to string when
setting it as environment variable.
Remove code fork for legacy runner, reorganize imports and move a
pylint-silencing snippet to the top of the file. The rest of the code
was just unindented.
In order to python system tests, pytest (runner) has to be used
directly. This makes it possible to simplify the pytest runner and make
its behavior simpler and easier to extend.
The legacy runner can still be used to run shell system tests.
The legacy runner no longer uses make check. Ensure the legacy runner
script doesn't interact with that automake target in any way. The legacy
runner script remains available to execute the legacy runner, but there
is no out-of-the box support for running tests in parallel. Other tools
such as xargs can be utilized for that.
Invoking pytest directly provides a better formatted output and more
flexibility. However, it's prudent to verify that `make check` keeps
working as expected. Use it in the OpenBSD job which isn't executed as
frequently and its output is of least concern.
Pytest provides JUnit output and uses different exit codes from
Automake. Use the conversion script to interpret the JUnit test results
from python rather than relying on the status code.
It's important to parse the JUnit result file rather than relying on the
exit code from pytest, which has a different meaning. Include a .trs test
result for each test case and set an exit code which is most appropriate
as the aggregate result (e.g. it will be set to 77 (SKIP) if there's at
least one test case that was skipped).
Dependencies for these tests are already checked in prereq.sh - if the
dependencies are missing, these tests will be skipped. The extra
dependency check in Makefile.am is extraneous and only applied for the
legacy test runner.
To allow concurrent invocations of pytest, it is necessary to assign
ports properly to avoid conflicts. In order to do that, pytest needs to
know a complete list of all test modules.
When pytest is invoked from run.sh, the current working directory is the
system test directory. To properly detect other tests, the conftest.py
has to look in the bin/tests/system directory, rather than the current
working directory.
When adding a new system test, it might easy to forget to add the
required files for the pytest runner or break a naming convention. Add
danger checks to cover these cases.
To conform with the expected naming convention, the pytest glue file for
the `allow-query` test should use underscore as the word separator in
the python file name: allow-query/tests_sh_allow_query.py
The _common directory is a special case directory which contains shared
files for other system test directories. Make sure it's tracked in git
and not deleted during temporary directory cleanup.
The old name "common" clashes with the convention of system test
directory naming. It appears as a system test directory, but it only
contains helper files.
To reduce confusion and to allow automatic detection of issues with
possibly missing test files, rename the helper directory to "_common".
The leading underscore indicates the directory is different and the its
name can no longer be confused with regular system test directories.
When iterating the table, we can't add new nodes to the hashmap because
we can't assure that we are not adding the new node before the iterator.
This also applies to rehashing - which might be triggered by both
isc_hashmap_add() and isc_hashmap_delete(), but not
isc_hashmap_iter_delcurrent_next().
When isc_hashmap_iter_delcurrent_next calls hashmap_delete_node
nodes from the front of the table could be added to the end of
the table resulting in them being returned twice. Detect when
this is happening and prevent those nodes being returned twice
buy reducing the effective size of the table by one each time
it happens.
Reusing TCP connections with dns_dispatch_gettcp() used linear linked
list to lookup existing outgoing TCP connections that could be reused.
Replace the linked list with per-loop cds_lfht hashtable to speedup the
lookups. We use cds_lfht because it allows non-unique node insertion
that we need to check for dispatches in different connection states.
Instead of high number of dispatches (4 * named_g_udpdisp)[1], make the
dispatches bound to threads and make dns_dispatchset_t create a dispatch
for each thread (event loop).
This required couple of other changes:
1. The dns_dispatch_createudp() must be called on loop, so the isc_tid()
is already initialized - changes to nsupdate and mdig were required.
2. The dns_requestmgr had only a single dispatch per v4 and v6. Instead
of using single dispatch, use dns_dispatchset_t for each protocol -
this is same as dns_resolver.
Looking up unique message ID in the dns_dispatch has been using custom
hash tables. Rewrite the custom hashtable to use cds_lfht API, removing
one extra lock in the cold-cache resolver hot path.
Refactor isc_hashmap to allow custom matching functions. This allows us
to have better tailored keys that don't require fixed uint8_t arrays,
but can be composed of more fields from the stored data structure.
