[bugfix] Fix Slurm job array completion detection

[dingp]

Summary

• recognize compact Slurm array IDs containing lists, ranges, steps and percent throttles
• track the requested array task count from the --array option
• keep an array nonterminal when a scheduler snapshot accounts for fewer tasks than requested
• add scheduler regression tests for incomplete and compact sacct responses

Background

Slurm can report a pending throttled array element as 56502501_[1%1]|PENDING. The previous job-ID pattern did not match that form. If sacct simultaneously reported an earlier element as completed, ReFrame could treat the array as complete and merge output before the remaining element ran.

Testing

• venv/bin/python test_reframe.py -q unittests/test_schedulers.py -k "slurm_count_array_tasks or slurm_array_jobid_pattern or slurm_poll_job_array" — 12 passed, 27 skipped
• venv/bin/python test_reframe.py -q unittests/test_schedulers.py — 166 passed, 226 skipped

[Copilot]

Pull request overview

This PR fixes ReFrame’s Slurm job-array completion detection when Slurm reports compact array element IDs (including lists/ranges/steps and percent throttles). It ensures a job array is not treated as complete (and its outputs merged) until the scheduler snapshot accounts for all tasks requested via --array.

Changes:

• Extend Slurm job-id matching to recognize compact array element expressions (e.g., 123_[1,3-7:2%2]).
• Track the requested array task count from the --array option and keep arrays nonterminal when fewer tasks are observed in scheduler snapshots.
• Add unit tests covering array task counting, job-id pattern matching, and polling behavior with incomplete/compact sacct outputs.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
reframe/core/schedulers/slurm.py	Adds array-expression task counting, broadens Slurm array job-id pattern matching, and prevents premature completion when fewer tasks are observed than requested.
unittests/test_schedulers.py	Adds regression tests for array task counting, compact job-id patterns, and polling behavior with incomplete/compact sacct responses.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[codecov]

Codecov Report

❌ Patch coverage is 95.00000% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 91.94%. Comparing base (780c2c4) to head (ac6e752).

Files with missing lines	Patch %	Lines
reframe/core/schedulers/slurm.py	95.00%	2 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #3707      +/-   ##
===========================================
- Coverage    91.97%   91.94%   -0.03%     
===========================================
  Files           62       62              
  Lines        13788    13825      +37     
===========================================
+ Hits         12681    12712      +31     
- Misses        1107     1113       +6     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[vkarak]

@dingp Thanks for the PR. Can you elaborate a bit on the problem with some more examples and the scenarios that the current implementation fails? A minimal reproducer test submitting a job array would be rather helpful in order to understand and evaluate better the fix.

[vkarak]

@dingp Thanks for the PR. Can you elaborate a bit on the problem with some more examples and the scenarios that the current implementation fails? A minimal reproducer test submitting a job array would be rather helpful in order to understand and evaluate better the fix.

The Slurm backend is crucial for most of the users, so we need to have a clear understanding of what exactly this PR fixes in order to avoid breaking the backend.