Prevent silent integer overflows in arithmetic operators

[rostan-t]

Category:

Bug fix (non-breaking change which fixes an issue)

Description:

Currently, arithmetic operators pull constant integers as 32 bits signed ints. As a result, silent overflows can happen, producing wrong results without any warning nor error.

The Python binding forOpSpec::AddArg is already registered as int64_t:

DALI/dali/python/backend_impl.cc

Line 3414 in 3e40f6d

DALI_OPSPEC_ADDARG(int64_t)

And the argument storage uses int64_t for all integral types anyway:

DALI/dali/pipeline/operator/argument.h

Lines 32 to 37 in 3e40f6d

	template <typename T>
	struct argument_storage {
	using type = std::conditional_t<
	std::is_integral<T>::value || std::is_enum<T>::value,
	int64_t, T>;
	};

So we can pull the integer constants as 64 bit integers and error out if they don't fit their target dtype.

Additional information:

This allows arithmetic operators to consume more than 32 bit signed integers but throws an exception instead of silently overflowing.

Affected modules and functionalities:

Integral constants in arithmetic operators.

Key points relevant for the review:

Tests:

• Existing tests apply
• New tests added
  • Python tests
  • GTests
  • Benchmark
  • Other
• N/A

Checklist

Documentation

• Existing documentation applies
• Documentation updated
  • Docstring
  • Doxygen
  • RST
  • Jupyter
  • Other
• N/A

DALI team only

Requirements

• Implements new requirements
• Affects existing requirements
• N/A

REQ IDs: N/A

JIRA TASK: N/A

[greptile-apps]

Confidence Score: 5/5

The PR appears safe to merge, with no outstanding correctness or repository-rule findings in the scoring set.

Summary

This PR prevents silent narrowing of integral constants in arithmetic expressions by transporting them as 64-bit integers and validating them against their explicitly selected destination types.

• Changes arithmetic-expression integer argument storage from 32-bit to 64-bit.
• Rejects out-of-range integral constants with an OverflowError that preserves operator context.
• Documents the transport and promotion behavior for Python constants.
• Adds boundary and overflow-rejection tests for signed and unsigned constants.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart LR
  A[Python integral constant] --> B[OpSpec integer_constants as int64]
  B --> C[ConstantStorage]
  C --> D{Fits selected dtype?}
  D -->|Yes| E[Cast and store constant]
  D -->|No| F[Throw OverflowError with operator context]

Loading

_{Reviews (5) · Last reviewed commit: "Update documentation to reflect that ari..."}

[rostan-t]

!build

[dali-automaton]

CI MESSAGE: [67521712]: BUILD STARTED

[dali-automaton]

CI MESSAGE: [67521712]: BUILD FAILED

[dali-automaton]

CI MESSAGE: [67521712]: BUILD PASSED

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[rostan-t]

!build

[dali-automaton]

CI MESSAGE: [67797265]: BUILD STARTED

[dali-automaton]

CI MESSAGE: [67797265]: BUILD FAILED

[rostan-t]

!build

[dali-automaton]

CI MESSAGE: [67967537]: BUILD STARTED

[dali-automaton]

CI MESSAGE: [67967537]: BUILD PASSED

[mzient]

I think this works by accident (kind of) - please also change the argument type in arithmetic.cc to DALI_INT64_VEC after #6486 is merged.

[rostan-t]

Done.

[rostan-t]

!build

[dali-automaton]

CI MESSAGE: [68224735]: BUILD STARTED

[dali-automaton]

CI MESSAGE: [68224735]: BUILD PASSED

[jantonguirao]

LGTM. Confirmed the earlier review concerns (int64 storage, overflow check semantics, UINT32 boundary test fix, CI failure) are all resolved, and CI is green.