Sound Static Data Race Verification for C: Is the Race Lost?
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Sound static data race freedom verification has been a long-standing challenge in the field of programming languages. While actively researched a decade ago, most practical data race detection tools have since abandoned soundness. Is sound static race freedom verification for real-world C programs a lost cause? In this work, we investigate the obstacles to making significant progress in automated race freedom verification. We selected a benchmark suite of real-world programs and, as our primary contribution, extracted a set of coding idioms that represent fundamental barriers to verification. We expressed these idioms as micro-benchmarks and contributed them as evaluation tasks for the International Competition on Software Verification, SV-COMP. To understand the current state, we measure how sound automated verification tools competing in SV-COMP perform on these idioms and also when used out of the box on the real-world programs. For 8 of the 20 coding idioms, there does exist an automated race freedom verifier that can verify it; however, we also found significant unsoundness in leading verifiers, including Goblint and Deagle. Five of the seven tools failed to return any result on any real-world benchmarks under our chosen resource limitations, with the remaining two tools verifying race freedom for 2 of the 18 programs and crashing or returning inconclusive results on the others. We thus show that state-of-the-art verifiers have both superficial and fundamental barriers to correctly analyzing real-world programs. These barriers constitute the open problems that must be solved to make progress on automated static data race freedom verification.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it