reframe-hpc/reframe

[2026-09-08T16:05:03.799945] debug: reframe: Loading user configuration [2026-09-08T16:05:03.800691] debug: reframe: Loading the builtin configuration [2026-09-08T16:05:03.800969] debug: reframe: Loading configuration file: '/gpfs/home5/satishk/projects/eessi_reframe/settings_example.py' [2026-09-08T16:05:03.823138] debug: reframe: '16_nodes' does not match '^[a-zA-Z_](?:[a-zA-Z0-9_-])*$' Failed validating 'pattern' in schema['properties']['systems']['items']['properties']['partitions']['items']['properties']['features']['items']: {'type': 'string', 'pattern': '^[a-zA-Z_](?:[a-zA-Z0-9_-])*$'} On instance['systems'][0]['partitions'][3]['features'][13]: '16_nodes' [2026-09-08T16:05:03.823379] error: reframe: �[31mERROR: failed to load configuration: could not validate configuration files: `<builtin>`, `/gpfs/home5/satishk/projects/eessi_reframe/settings_example.py`: '16_nodes' does not match '^[a-zA-Z_](?:[a-zA-Z0-9_-])*$'�[0m [2026-09-08T16:05:03.823557] info: reframe: Log

@vkarak - let me know if this makes sense, I quickly gathered below from an external cluster: Description ReFrame can hang / wait indefinitely when a Slurm job fails during node startup/prolog handling and Slurm accounting records the top-level job as REQUEUED, even though the job is no longer active. In this case, the job hit a prolog/startup issue on a GPU node. The job then entered CG while epilog handling was still active, and I cancelled it manually. After cancellation: squeue no longer showed the job scontrol show job reported the job id as invalid sacct continued to report the top-level job as REQUEUED ReFrame continued polling ? Environment ReFrame version: master branch f1744f0 Scheduler: Slurm Launcher: srun Test type: RunOnlyRegressionTest Test command: simple nvidia-smi query Logs $ /path/to/reframe -C sites/site-config.py -c checks/nvidia_smi.py --system cluster:gpu_preflight -r [2026-07-17T12:45:56.983732] info: reframe: [ReFrame Setup] [2026-07-17T12:45:56.983968]

I have a module whose behavior I can modify by setting an environment variable before loading it. This module loads an environment, that again provides other modules. I.e. interactively, I'd do: export X=foo module load Y/1.1 # the load behaviour is affected by the value of X module avail bar # bar is available in this environment (but only if X=foo) I'd like reframe.utility.find_modules to be able to find bar. I've noticed that I cannot simply specify environment.env_vars because those only get set after reframe.core.runtime.loadenv has already called the module commands. So, I figured I'll just leverage prepare_cmds instead: 'environments': [ { 'prepare_cmds': ['export X=foo'], 'modules': ['Y/1.1'] } Alas, this didn't work. Looking at the code, I can see why: the prepare_cmds are appended to the returned commands here, but they are not executed when loadenv() is called. According to my AI friend, they are only emitted to the job script that reframe stages