[patrick-immometrica]

Summary

peek_filelike_length() takes os.fstat().st_size when the object has a fileno(). That's wrong for anything where read() doesn't return the raw file bytes:

• gzip.open(path, "rb") reports the compressed size
• a pipe (Popen(...).stdout, sys.stdin.buffer) reports 0

So files={"f": gzip.open(...)} sends a too small Content-Length and fails with LocalProtocolError: Too much data for declared Content-Length.

This drops the fstat branch and keeps the tell()/seek() fallback, the same thing #1214 did for content=. Gzip files now get the right length, pipes go out chunked. Regular files are unchanged.

Checklist

• I understand that this PR may be closed in case there was no previous discussion. (This doesn't apply to typos!)
• I've added a test for each change that was introduced, and I tried as much as possible to make a single atomic change.
• I've updated the documentation accordingly.

[cubic-dev-ai]

All reported issues were addressed across 2 files

<sub>Reply to a comment to ask cubic a question or push back. It learns from your replies.

View guided diff | Re-trigger cubic</sub>

[codspeed]

Merging this PR will not alter performance

✅ 17 untouched benchmarks
⏩ 8 skipped benchmarks¹

---

_{Comparing patrick-immometrica:multipart-file-length (6efb783) with main (f4f65c1)}

Footnotes

1. 8 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩