The concurrency_test is much more stable than it used to be back in
0.9.3, but I agree there are still some issues left.  It just caused
an Appveyor failure, and will very occasionally cause a Travis
failure.  I don't know what the root cause is right now.

- Jim

On Sat, May 11, 2019 at 1:06 PM Jens Geyer <[email protected]> wrote:
>
> Not sure whether we need to be concerned or if the test may just be flawed:
>
> [00:16:23] 451: ThreadManager benchmark tests...
> [00:16:23] 451:   ThreadManager load test: worker count: 2 task count: 2000 
> delay: 5
> [00:16:23] 451:     loaded 2000 tasks to execute
> [00:16:23] 451:     activeCount = 2000
> [00:16:38] 451:    first start: 1122613 Last end: 1138293 min: 6ms max: 21ms 
> average: 15.6485ms
> [00:16:38] 451:    Success! expected time: 5000ms elapsed time: 15680ms
> [00:16:38] 451:   ThreadManager load test: worker count: 8 task count: 8000 
> delay: 5
> [00:16:38] 451:     loaded 8000 tasks to execute
> [00:16:38] 451:     activeCount = 7992
> [00:16:54] 451: Assertion failed: delta > 0, file 
> c:\projects\thrift\lib\cpp\test\concurrency\ThreadManagerTests.h, line 162
> the code tests the exec time for all tasks:
>
>     for (auto ix = tasks.begin(); ix != tasks.end(); ix++) {
>       shared_ptr<ThreadManagerTests::Task> task = *ix;
>       int64_t delta = task->_endTime - task->_startTime;
>       assert(  > 0);    // <<<<<<<<<<<<<<
>
> Thoughts?
>
> https://ci.appveyor.com/project/ApacheSoftwareFoundation/thrift/builds/24472298/job/19m0dvxkusbwqwn6
>
> Have fun,
> JensG
>

Reply via email to