[SERVER-19309] DBtest fails ThreadedTests::Slack<mongo::SimpleRWLock, mongo::SimpleRWLock::Exclusive> Created: 07/Jul/15 Updated: 25/Aug/15 Resolved: 25/Aug/15 |
|
| Status: | Closed |
| Project: | Core Server |
| Component/s: | Concurrency, Testing Infrastructure |
| Affects Version/s: | None |
| Fix Version/s: | None |
| Type: | Bug | Priority: | Major - P3 |
| Reporter: | Marek Skalický | Assignee: | Sam Kleinman (Inactive) |
| Resolution: | Incomplete | Votes: | 0 |
| Labels: | None | ||
| Remaining Estimate: | Not Specified | ||
| Time Spent: | Not Specified | ||
| Original Estimate: | Not Specified | ||
| Operating System: | ALL | ||||||||||||||||||
| Steps To Reproduce: | It is build by:
Where %{?optflags} = "-O2 -g -pipe -Wall -Werror=format-security -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector-strong --param=ssp-buffer-size=4 -grecord-gcc-switches -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -m64 -mtune=generic" and %{?__global_ldflags} = "-Wl,-z,relro -specs=/usr/lib/rpm/redhat/redhat-hardened-ld". Because it is not failing in all cases, it is necessary to run dbtest more times... Script to reproduce:
|
||||||||||||||||||
| Participants: |
| Description |
|
When I run "./dbtest --dbpath ./var/dbtest" it fails with "dbtest: /usr/include/boost/thread/pthread/condition_variable_fwd.hpp:86: boost::condition_variable::~condition_variable(): Assertion `!ret' failed."
See this bug - https://jira.mongodb.org/browse/SERVER-18906 I have a wrong assumption that this is caused by --dbpath handling. It could be some race condition problem. It is not failing in every run. All tests in threading suite pass fine, only in some cases it fails. |
| Comments |
| Comment by Sam Kleinman (Inactive) [ 09/Jul/15 ] | ||||||||||||
|
Thanks for this update. Can you also provide more information about the C++ standard library and version? At the moment I'm most interested in reproducing recent master versions built with using all of the bundled (vendored third_party libraries.) Once we can isolate and reproduce the issue there we can move to fix the issue. If it's only reproducible with the more recent boost we can test with the --use-system-boost to try and isolate that variable. The dbTest suites run as part of the compile step, which is the first step (box) in each column in the evergreen display, and you can use that display to find commits that have passed on every other platform to test on Fedora Rawhide. I will leave this issue open while we wait for you to get back online. Take care! Cheers, | ||||||||||||
| Comment by Marek Skalický [ 09/Jul/15 ] | ||||||||||||
|
boost - 1.57 -> I tried to reproduce. And I was successful with 3.1.5 and also 3.1.6 (for example for bundled libs I have to run dbtest 200 times ~ the script 2 times) Also some tests failed with
I will try to provide more test details. Thanks, | ||||||||||||
| Comment by Sam Kleinman (Inactive) [ 08/Jul/15 ] | ||||||||||||
|
Thanks for getting back to me. I'm particularly interested in the versions of the packages that your build is using rather than using the vendorized packages, which is probably just Boost in this case. The version of libc that you're using may also be useful just as a data point. I want to try and isolate the cause of this error a bit more, and my own attempts to reproduce the problem using your script have not yet been successful. Could you attempt two different tests to see if you can reproduce the error:
Cheers, | ||||||||||||
| Comment by Marek Skalický [ 08/Jul/15 ] | ||||||||||||
|
Thank you. 1. Sorry, I was wrong. It was built with wiredtiger=on... I will try to build it with wiredtiger=off to check if it is failing for me.
5. I tried to run dbtest with MongoDB 3.0.4, also 3.1.5 failed with this error. I havent tried earlier versions. | ||||||||||||
| Comment by Sam Kleinman (Inactive) [ 07/Jul/15 ] | ||||||||||||
|
I'm attempting to reproduce this right now using the latest master, and I will keep you posted with my progress. I have a couple of additional questions and clarifications while I work on the reproduction.
Regards, |