Guard base64 data URL image decoding against unbounded memory allocation

[Copilot]

persist_data_image_to_temp_file decoded base64 payloads with no size checks, leaving it open to OOM from large or malicious inline images.

Changes

• Pre-decode guard: Estimates decoded size via encoded.len() * 3 / 4 and rejects before allocating if it exceeds URL_IMAGE_MAX_BYTES (8 MB)
• Post-decode guard: Checks actual bytes.len() after decoding as a safety net against edge cases in the estimate

let estimated_len = encoded.len().saturating_mul(3) / 4;
if estimated_len > URL_IMAGE_MAX_BYTES {
    return None;
}
let bytes = base64::engine::general_purpose::STANDARD
    .decode(encoded)
    .ok()?;
if bytes.len() > URL_IMAGE_MAX_BYTES {
    return None;
}

Both guards reuse the existing URL_IMAGE_MAX_BYTES constant already defined in scope.

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.

---

Summary by cubic

Add pre- and post-decode byte caps for data:image/*;base64, in src-tauri/src/shared/codex_core.rs to reduce memory spikes in heavy threads. Oversized inline images are rejected early to avoid large allocations.

• Bug Fixes
  • Estimate decoded size with encoded.len().saturating_mul(3)/4 and skip if above URL_IMAGE_MAX_BYTES.
  • After decoding, check bytes.len() against URL_IMAGE_MAX_BYTES; return None if exceeded.

^{Written for commit a279298. Summary will update on new commits.}

[cubic-dev-ai]

No issues found across 1 file