Repository navigation
http_client test failing on mac with a segfault. #903
Description
Making a macports portfile, on 2.10.6, testing has following output:
Running tests...
/opt/local/bin/ctest --force-new-ctest-process
Test project /opt/local/var/macports/build/_opt_portfiles_www_cpprestsdk/cpprestsdk/work/build
Start 1: httpclient_test
1/7 Test #1: httpclient_test ..................***Exception: SegFault 29.16 sec
All other tests pass.
Running those tests manually via test_runner, I get this:
Test case request_helper_tests:overwrite_user_agent PASSED
Starting test case request_stream_tests:set_body_stream_1...
/opt/local/var/macports/build/_opt_portfiles_www_cpprestsdk/cpprestsdk/work/cpprestsdk-2.10.6/Release/tests/functional/http/client/request_stream_tests.cpp:96: error: Failure in set_body_stream_1: Unhandled exception: No such file or directory: No such file or directory FAILED
Test case request_stream_tests:set_body_stream_1 FAILED
Starting test case request_stream_tests:set_body_stream_2...
/opt/local/var/macports/build/_opt_portfiles_www_cpprestsdk/cpprestsdk/work/cpprestsdk-2.10.6/Release/tests/functional/http/client/request_stream_tests.cpp:123: error: Failure in set_body_stream_2: Unhandled exception: No such file or directory: No such file or directory FAILED
Test case request_stream_tests:set_body_stream_2 FAILED
Starting test case request_stream_tests:without_content_length_1...
/opt/local/var/macports/build/_opt_portfiles_www_cpprestsdk/cpprestsdk/work/cpprestsdk-2.10.6/Release/tests/functional/http/client/request_stream_tests.cpp:193: error: Failure in without_content_length_1: Unhandled exception: No such file or directory: No such file or directory FAILED
libc++abi.dylib: terminating
Trace/BPT trap: 5
Activity
BillyONeal commentedon Oct 18, 2018
Does it still repro for you with master? I just merged a fix for a bunch of flaky test problems.
Lord-Kamina commentedon Oct 18, 2018
I will check tonight but I'd say probably yes, from a quick look at your commit, it doesn't seem to touch the files mentioned in the log I posted.
BillyONeal commentedon Oct 18, 2018
Right, that's because there were product bugs; the tests were catching legitimate races and memory corruption bugs in cpprestsdk proper.
Lord-Kamina commentedon Oct 18, 2018
Will try them again tonight when I get off work then.
Lord-Kamina commentedon Oct 19, 2018
95409c5 is passing all tests now on macOS 10.14
BillyONeal commentedon Oct 19, 2018
Unfortunately the tests remain a bit flaky because the experimental::asio_http_listener is used to test the client -- and that is full of race conditions and memory corruption bugs :(. The client is the thing we actually care about and I haven't found memory corruption bugs in that yet at least.
Since the specific easily reproducable problem is gone I'll close this for now.