DIALS meeting 2020-11-11
Azure caching is completely broken, we’ve disabled caching and multi-python builds for now. Pull requests that aren’t based on a recent
master may have failing builds. Merge
master into them to fix.
- Shouldn’t be controversial.
- AP: GW to test locally and approve
- RG: Although there are no specific objections there is an appreciation that this will break everything.
- Specific use case is anything with large number of experiments and you then want to go back to indexing later
- Analogy: In indexing we don’t throw away unindexed reflections
- DGW: I think it will work. It probably will break things, but those can be changed. In principle this sounds reasonable.
- Outcome: Cautious support for the principle.
- ND: If we add nproc=auto then we bog down the test machine
- MG: If we use nproc=1 in tests we run a different code path to users.
- GW: Should possibly set nproc=2 in tests to avoid the code path issue.
- RG: It will make dials easier to run for the user
- Outcome: Merge, if there are any problems then we’ll address that in the future.
dxtbx Nexus format support
- Nexus files generated at Diamond do follow the Nexus spec.
- dxtbx makes parsing correct nexus files hard, as there are too many different nearlybutnotquite-formats that try to jump in with broken geometry assumptions etc.
- AP: BW to follow up
- https://github.com/dials/dials/pull/1398#issuecomment-690123149 last on this topic - guess probably couple hours work needed?
Adding GEMMI as dependency of DIALS
Would be useful for xia2 mmcif stuff?
- MG: Would hold back on this until the conda-forge package is fixed: https://github.com/conda-forge/gemmi-feedstock/issues/20
- AP: DW to follow up
Not discussed in this meeting
- @James B-E you had something to pick up re: anomalous signal (presence or absence thereof?)
- Outcomes from last collaboration call with RJR etc.