MaxTimeoutTest.ContentStream has started to fail on x86_64

[opoplawski]

The Fedora cpp-httplib package has started to fail building on x86_64: https://koschei.fedoraproject.org/package/cpp-httplib

The failure is:

310/312 Test #310: MaxTimeoutTest.ContentStream .........................................***Failed    6.13 sec
Running main() from /builddir/build/BUILD/gtest-1.15.2-build/googletest-1.15.2/googletest/src/gtest_main.cc
Note: Google Test filter = MaxTimeoutTest.ContentStream
[==========] Running 1 test from 1 test suite.
[----------] Global test environment set-up.
[----------] 1 test from MaxTimeoutTest
[ RUN      ] MaxTimeoutTest.ContentStream
/builddir/build/BUILD/cpp-httplib-0.19.0-build/cpp-httplib-0.19.0/test/test.cc:8266: Failure
Value of: success
  Actual: true
Expected: false
[  FAILED  ] MaxTimeoutTest.ContentStream (6004 ms)
[----------] 1 test from MaxTimeoutTest (6004 ms total)
[----------] Global test environment tear-down
[==========] 1 test from 1 test suite ran. (6004 ms total)
[  PASSED  ] 0 tests.
[  FAILED  ] 1 test, listed below:
[  FAILED  ] MaxTimeoutTest.ContentStream
 1 FAILED TEST

This still fails with 0.20.0. There are some slight changes in various dependencies but nothing is standing out to me. Any ideas?

[yhirose]

@opoplawski thanks for the report. Could you please provide the error log with 0.20.0 as well? Thanks!

[opoplawski]

Running main() from /builddir/build/BUILD/gtest-1.15.2-build/googletest-1.15.2/googletest/src/gtest_main.cc
Note: Google Test filter = MaxTimeoutTest.ContentStream
[==========] Running 1 test from 1 test suite.
[----------] Global test environment set-up.
[----------] 1 test from MaxTimeoutTest
[ RUN      ] MaxTimeoutTest.ContentStream
/home/orion/fedora/cpp-httplib/cpp-httplib-0.20.0-build/cpp-httplib-0.20.0/test/test.cc:8528: Failure
Value of: success
  Actual: true
Expected: false

/home/orion/fedora/cpp-httplib/cpp-httplib-0.20.0-build/cpp-httplib-0.20.0/test/test.cc:8589: Failure
Value of: res
  Actual: true
Expected: false

[  FAILED  ] MaxTimeoutTest.ContentStream (7006 ms)
[----------] 1 test from MaxTimeoutTest (7006 ms total)

[----------] Global test environment tear-down
[==========] 1 test from 1 test suite ran. (7006 ms total)
[  PASSED  ] 0 tests.
[  FAILED  ] 1 test, listed below:
[  FAILED  ] MaxTimeoutTest.ContentStream

 1 FAILED TEST

[opoplawski]

Another run yielded:

[  FAILED  ] MaxTimeoutTest.ContentStream (5004 ms)
[----------] 1 test from MaxTimeoutTest (5004 ms total)

[----------] Global test environment tear-down
[==========] 1 test from 1 test suite ran. (5004 ms total)

[opoplawski]

This seems to be an intermittent failure.

[DarthGandalf]

Same here (Gentoo, 0.20.0):

[----------] 2 tests from MaxTimeoutTest
[ RUN      ] MaxTimeoutTest.ContentStream
test.cc:8528: Failure
Value of: success
  Actual: true
Expected: false
test.cc:8589: Failure
Value of: res
  Actual: true
Expected: false
[  FAILED  ] MaxTimeoutTest.ContentStream (7003 ms)
[ RUN      ] MaxTimeoutTest.ContentStreamSSL
test.cc:8507: Failure
Value of: success
  Actual: true
Expected: false
test.cc:8574: Failure
Value of: res
  Actual: true
Expected: false
[  FAILED  ] MaxTimeoutTest.ContentStreamSSL (5007 ms)
[----------] 2 tests from MaxTimeoutTest (12010 ms total)

[falbrechtskirchinger]

Well, we kind of knew that. See #2059.

I need to measure SSL_shutdown() again. When I rewrote it to account for max timeout, it measured 0ms as expected, though the test still failed because the timeout was exceeded elsewhere in the code (by hundreds of milliseconds!).

Edit: To clarify: I implemented two versions of SSL shutdown that honoured max timeout, but neither made it into #2059. The fix was quite invasive, and I left it up to @yhirose to decide how to proceed. Neither one of us followed up, so the issue remains unresolved.

[yhirose]

@opoplawski, @DarthGandalf could you please test with the latest httplib.h on the master branch? Thanks!

[DarthGandalf]

Works for me now