Daniel Russel wrote: > Here are some changes to all spheres nonbonded list to fix a bug that > Frido found as well as make things faster. The nonbonded lists now have > checks built into them so asserts should fail if anything goes wrong. > Note that these checks are very expensive for large numbers of particles > (quadratic) and so any real runs with IMP should use builds with NDEBUG > defined. We may want to define macros like IMP_CHECK_EXPENSIVE to > provide finer grain control over debugging checks.
Note: you can say "scons release=true" to turn off most debugging checks. But the nightly builds use the default (release=false) since we need all the testing we can get right now.
BTW, the Frido-sphere test doesn't seem to contain any assertions, so I'm not sure what it's supposed to test (although the particle coordinates seem rather precise). Is it supposed to crash without this patch? (It seemed to work fine on synth without it.) If so, a comment in the test case saying something like "causes a segfault with r456 or older due to a bug in foo" would prevent future people from deleting the seemingly useless test.
Ben