[samirhvbr]

Fixes #1620.

Cloud downloads stream their responses. If a server returns an HTTP error and its JSON or text body is truncated, reading that body currently lets requests.exceptions.ChunkedEncodingError escape from the CLI. For example, scans report --output report.pdf --json produces a traceback instead of a JSON error on an interrupted 503 response.

Convert Requests body-read failures in http.check to the existing CloudTransportError. The runner can then return exit 1 through its normal output handling: JSON for file downloads and stderr for redirected binary output. The response still closes and existing files remain untouched.

Regression tests run through the real CLI and Requests/urllib3 body readers with an incomplete response, mocking only the HTTP request. All four cases fail on unchanged main and pass with this fix: JSON/text error bodies, each sent to a file or raw stdout.

Validation on Python 3.12.14 with frozen dependencies and an isolated HOME:

• All seven cloud test modules: 325 passed.
• make check-all: Ruff formatting/lint and full Mypy pass; Pyright stops the target with 941 pre-existing errors. The unchanged base and this branch have identical diagnostic sets (zero additions/removals); http.py has no diagnostics.
• Full Bandit run separately: passed.
• git diff --check: passed.

No live managed service, LLM, payment, or scan was used. Network behavior is covered by local synthetic responses; remote CI remains to be run.

[greptile-apps]

Confidence Score: 5/5

[Medium impact] The PR appears safe to merge; no actionable issues were found.

Summary

Converts Requests body-read failures in check into CloudTransportError, so interrupted download error bodies reach the CLI’s normal error output.

• Unreadable cloud response bodies now become normal CLI errors.

_{Reviews (1) · Last reviewed commit: "fix(cloud): handle interrupted download ..." · Reviewed by Greptile}