checkpointing incorrect for simulations with . in name
Bug #1076967 reported by
Samuel Parkinson
This bug affects 1 person
| Affects | Status | Importance | Assigned to | Milestone | |
|---|---|---|---|---|---|
| Fluidity |
Confirmed
|
Medium
|
Simon Mouradian | ||
Bug Description
When running in parallel with a simulation that contains a . in the simulation name i.e. test.01 the checkpoint files are named in a way that isn't recognised when trying to rerun from a checkpoint.
As an example, the filepath for the vtu files become:
test.01_
instead of what is expected:
test.01_
where ** is the process number
Which results in not being able to rerun from the checkpoint without first renaming all of the files.
| Changed in fluidity: | |
| importance: | Undecided → Medium |
| Changed in fluidity: | |
| assignee: | nobody → Simon Mouradian (mouradian) |
Revision history for this message
| Simon Mouradian (mouradian) wrote | #1 |
| Changed in fluidity: | |
| status: | New → Confirmed |
To post a comment you must log in.
I confirm this is indeed a bug. Attached is an example which fails. To reproduce, try the following commands:
mpiexec -n 2 fluidity -v2 -l ismip_config.flml
mpiexec -n 2 bin/fluidity -v2 -l ismip.hom_
Changing the 'simulation_name' in the .flml to one without a dot (.) fixes the problem.