This changeset makes it possible to build the Second Life viewer using
Python 3. It is designed to be used with an equivalent Autobuild branch
so that a developer can compile without needing Python 2 on their
machine.
Breaking change: Python 2 support ending
Rather than supporting two versions of Python, including one that was
discontinued at the beginning of the year, this branch focuses on
pouring future effort into Python 3 only. As a result, scripts do not
need to be backwards compatible. This means that build environments,
be they on personal computers and on build agents, need to have a
compatible interpreter.
Notes
- SLVersionChecker will still use Python 2 on macOS
- Fixed the message template url used by template_verifier.py
This is the function in indra/llmessage/tests/testrunner.py that iterates
through ports in a specified range, looking for an available one. Other
platforms understand a specification of port 0 to mean: "You pick one. I'll
just use whichever one you picked."
Instead of having testrunner.run()'s caller pass a Thread object on which to
run the caller's server instance's serve_forever() method, just pass the
server instance. testrunner.run() now constructs the Thread. This API change
allows run() to also call shutdown() on the server instance when done, and
then join() the Thread.
The hope is that this will avoid the Python runtime forcing the process
termination code to 1 due to forcibly killing the daemon thread still running
serve_forever().
While at it, eliminate calls to testrunner.freeport() -- just make the runtime
pick a suitable port instead.
Over the years we've skipped more and more of the tests in
llhttpclient_test.cpp (nee llhttpclient_tut.cpp) because they've relied on
particular behaviors from Internet sites not under our control.
We skipped a test that fetches llsd+xml from secondlife.com because
"secondlife.com is not reliable enough for unit tests."
We skipped a test that tries to observe a failure with "http://www.invalid"
because some local ISPs turn "no such domain" DNS errors into valid pages
offering the requester to buy the specified domain name.
Today we've had to skip tests attempting to contact "http://www.google.com"
for reasons we haven't yet diagnosed, but that probably have to do with
Google's IPv6 rollout.
Use local temp server test_llsdmessage_peer.py as the success destination,
eliminating DNS, Internet access and remote server behavior as failure modes.
Use idle localhost port for failure test.
Re-enable all skipped LLHTTPClient tests!
Re-enable on Windows!
In support of these tests, modify test_llsdmessage_peer.py:
Support HEAD as no-data variant of GET.
Change GET result dict to avoid resembling an error response -- confusing.
Make GET/POST return actual dict rather than undecorated string "success".
Because of that last, change llcapabilitylistener_test.cpp and
llsdmessage_test.cpp to extract "reply" key from response rather than
expecting response to be a string itself.
Turns out that BaseHTTPServer.HTTPServer turns on that flag by default, which
causes freeport() to fail (on Windows only?), happily instantiating multiple
servers on the same port. Change known instances, fix freeport() docstring to
highlight the issue. Add freeport() unit tests to verify expected behavior.
Make testrunner.py module interpret $INTEGRATION_TEST_VERBOSE environment
variable, setting module global VERBOSE. Enable/disable debug() output based
on that variable, defaulting to VERBOSE True. Add debug() output to
freeport(), including reporting exceptions.
Add debug() output to test_llsdmessage_peer.py, including normal
BaseHTTPRequestHandler output: when VERBOSE is set, don't suppress
log_request() or log_error() output.
Add C++ verbose() function to query $INTEGRATION_TEST_VERBOSE, broken out as
two functions so we only have to interpret the value once. Default to 'true'.
Move C++ commtest_data::getport(variable) function to global namespace, broken
out as two functions to cache the value. Report value received when verbose()
returns true.
In indra/llmessage/tests/testrunner.py, introduce new freeport() function to
try a caller-specified expression (such as instantiating an object that will
listen on a server port) with a range of candidate port numbers until the
expression produces a value instead of EADDRINUSE exception.
Change test_llsdmessage_peer.py and test_llxmlrpc_peer.py to use freeport() to
construct their server class inline BEFORE launching the thread that will run
it, then pass that server's serve_forever method to daemon thread. Also set
os.environ["PORT"] to selected environment variable before running subject
test program.
In indra/llmessage/tests/commtest.h, introduce commtest_data::getport() to
read port number from specified environment variable, throwing exception if
variable not set or non-numeric. Construct default LLHost from getport("PORT")
instead of hardcoded constant.
Change indra/newview/tests/llxmlrpclistener_test.cpp to use commtest_data::
getport("PORT") instead of hardcoded constant. Also use LLSD::with() rather
than older LLSD::insert() syntax.
HOWEVER -- I am irritated to discover that llxmlrpclistener_test IS NOT RUN or
even built by newview/CMakeLists.txt! It's not even commented out -- it's
entirely deleted! I am determined to restore this test. However, as it will
take some fiddling with new link-time dependencies, that will be a separate
commit.
Bennett Goble