Re: [AMBER] SIGSEVV error on test_at_serial for QMMM

From: David Case <>
Date: Fri, 5 Jul 2019 12:33:04 +0000

On Thu, Jul 04, 2019, Francesca Lønstad Bleken wrote:
>It is CENTOS7 and the following modules loaded
>And this is the configure for the serial version. (for mpi I then use -intelmpi, but these errors is before that)
>yes | ./configure -noX11 intel > 01a-config_intel.log
>make -j 16 > 01b-make_intel.log
>make install > 01c-makeinstall_intel.log
>make test > 01d-maketest_intel.log
>But this is the installation with the fewest errors in the make test I have seen...

I'm not sure what the final sentence means here: do you mean that other
options (e.g. to configure) lead to more errors?

I'm not sure how much test coverage we have for the 2019 Intel
compilers. Try this:

1. go to AMBERHOME/AmberTools/src, and edit the configure_python script.
Find the line that says MINICONDA_VERSION=4.5.12 and change that to
MINICONDA_VERSION=latest. That should fix the problem getting python
configured (but should be unrelated to the QMMM segfaults)

2. Try ./configure -debug -noX11 gnu; make install; make test. We test gnu
compilers a lot, and (as I said before) have not seen reports before of
the type of error you are reported. (Avoid the "-j 16" option to make:
it's possible that there is some bad race condition in the parallel
build that we've not seen.) The -debug option may give a little more
helpful messages when tests fail.

3. Assuming you still get errors, remind us again of exactly which tests
are failing, and post the tracebacks you see.


AMBER mailing list
Received on Fri Jul 05 2019 - 06:00:04 PDT
Custom Search