New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Some tests fail/hang on FreeBSD #29
Comments
This is the same issue as #18. I'm not sure test 89 running 20 mins makes sense. Which architecture? Which processor? If you're running with a single core and a slow processor it's very well possible the calculation takes a long time since the VV10 kernel takes a long time. |
Indeed, nothing freezes, only 17 failures:
|
A lot more tests fail with the parallel version:
|
Do all these test failures prevent erkale's normal use? |
What is the segfault caused by? You should find the core dumps under the tests/ directory of your object directory. The problems with the sequential version probably aren't a cause for worry; that might be just a compiler bug; but nothing should be segfaulting... |
It's a recurring (cyclic) exception:
|
Compile the code with proper debug flags please, this is not saying where the problem arises. It might be a problem in libcxxrt; the code works just fine with glibc. |
For the testcase This never gets actually printed. Then it throws another exception: This also is never printed. Same with So on and so forth. When I add my own printing line (printf or std::cout<<), these are also never printed for some reason. Eventually |
Right, since instead of running the test suite you're now running the program separately without having passed the necessary environment variables - it doesn't know where the basis sets reside. Set |
Oh no, I do supply these variables:
Only |
OK, I realize there were conflicting options in a few tests since I had changed the defaults at some point; this is fixed in #32. |
At least |
And that's a weird failure - it fails in the test of the electron momentum density transform, which works for everything else, however. There's a tiny numerical difference in <p^2>/2 vs T for H2; this has worked before perfectly. It shouldn't be a huge problem though. |
What's the thing with your segfaults? |
Revision 9f5593b has 10 failing tests (down from 17 before):
The problem also is that erkale doesn't process exceptions properly, so it is necessary to build with debug info and use debugger, that makes troubleshooting more difficult. |
"The problem also is that erkale doesn't process exceptions properly," Uhh.. what do you mean by this? Exceptions crash the program when they are meant to do so. Debugging always requires to build with debuginfo. All distributions build with it, and package it separately, so meaningful backtraces can be reported when a package is not working for some reason or another. This still does not explain your original problem report where all the tests segfaulted. |
All tests should work now, please confirm. |
To compare,
|
Please run ctest. |
---sequential failures---
---parallel failures---
|
OK so your BLAS is broken,. Not a problem in ERKALE. |
Wait, what about the failure of |
The error is basically that the checkpoint file is corrupt, and since the XRS program checks if a previous calculation exists it will use it, it probably hasn't run the calculation at all. |
It was indeed the problem in OpenBLAS that it limited the number of threads to 16. Now I have it increased to 128. But Also, this exception is thrown at rev. 0927d8c |
Run ctest like I told you to, since you're not running |
|
... and was the object directory clean like I asked you? Please follow the procedure I asked you yesterday. |
It is clean. This is the first test run there. |
Umm, that makes no sense - you have to run water_dimer_wb97x first, like the test suite does. Like I said. Please run |
These tests fail:
Then, this test doesn't finish in 20 mins:
FreeBSD 11.2 amd64
The text was updated successfully, but these errors were encountered: