Consider Calloc and Realloc for autotuners

[sim642]

While debugging #2030 (comment) I noticed that the mallocWrappers autotuner doesn't pick up on the wrapper of calloc. Despite the name, it makes sense to do so (and with realloc). The mallocWrapper analysis also already considers them.

After reviewing all pattern matchings for these allocation kinds, I also found the same being missing from the allocation loop unrolling autotuner.

TODO

• sv-benchmarks

[Copilot]

Pull request overview

This PR updates Goblint’s autotuning heuristics so they treat calloc and realloc similarly to malloc, aligning the autotuners with existing allocation-kind handling in the malloc-wrapper analysis.

Changes:

• Extend the “allocations-in-loops” autotuner to also detect calloc/realloc calls inside loops.
• Extend the malloc-wrapper autotuner to recognize wrappers around calloc/realloc in addition to malloc (by reusing wrapper-function classification logic).
• Add a clarifying TODO comment in c2poAnalysis about missing realloc handling.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 3 comments.

File	Description
src/autoTune.ml	Autotuners now consider Calloc/Realloc for loop allocation detection and malloc-wrapper detection.
src/analyses/wrapperFunctionAnalysis.ml	Exposes wrapper-classification plumbing to enable reuse by autotuners.
src/analyses/c2poAnalysis.ml	Comment-only change noting a potential missing Realloc case.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[sim642]

On sv-benchmarks with level01, 60s and 1GB this already has interesting effects: https://goblint.cs.ut.ee/results/340-all-level01-pr-2038-after/table-generator-cmp.diff.html#/table?filter=0(0*status*(status(notIn(FAILED)),category(notIn()))).
To summarize:

• 8 trues go to timeout/OOM
• 10 timeouts/OOMs go to true, most by significant margin
• 3 unknowns go to true
  • 1 of them is unsound, which suggests an underlying issue because more malloc wrappers should not have that effect on its own.

[michael-schwarz]

Looks reasonable!

[sim642]

• 1 of them is unsound, which suggests an underlying issue because more malloc wrappers should not have that effect on its own.

I looked into it and according to manual inspection the branches that became dead due to this change seemed to actually be dead. I then looked at SV-COMP 2026 results and found out that the task was invalidated. Turns out Mopsa already had an issue with this:
https://gitlab.com/sosy-lab/benchmarking/sv-benchmarks/-/blob/a45aa8f2eae1325c6c555c945ea721c023dde6f3/c/ldv-linux-4.2-rc1/linux-4.2-rc1.tar.xz-32_7a-drivers--pcmcia--pcmcia_rsrc.ko-entry_point.cil.out.yml. Apparently, I never updated sv-benchmarks on our server. So there should be no actual problem in Goblint.