tinyhttps 0.2.10

[Sunrisepeak]

send_stream kept the status line of a failed streaming request and dropped the reason — so a caller could report 400 Bad Request and never model_not_found. Reported and first fixed by @Cloud_Yun in mcpplibs/tinyhttps#11 and #12; landed with the review in mcpplibs/tinyhttps#14.

Three framing defects found while reviewing that change are fixed with it:

• A declared Content-Length is honoured by send_stream. It had no branch for one, so a non-chunked response was read until the connection closed — and on the library's defaults (keepAlive = true) the server does not close. Measured against a 135-byte error document: 9379 ms at an 8-second read timeout, against 1192 ms now. At the shipped default of 60000 that was a minute.
• A chunk size that does not parse is no longer a terminal chunk. parse_hex returned what it had accumulated on an unrecognised character and 0 for an empty line — which is what read_line returns on a timeout or a closed connection — so a truncated stream read as one that ended cleanly.
• Content-Length is rejected rather than salvaged. "abc" was 0 and indistinguishable from a real zero; "-1" was 1; "99999999999999999999" wrapped to 7766279631452241919.

Measured

http=200 bytes=29469
got  7ab1555ff465ae179579cc2d9328db11c847770f9a22148d563cd041dd8e4481   IDENTICAL
合法归档 ✓

Fetched back from the published CN URL, hashed against the GitHub tag tarball, and opened as an archive.

Locally: check_mirror_urls, check_package_name, check_cross_package_refs, check_platform_version_parity, check_duplicate_versions all pass. Appended after 0.2.9 in all three platform tables, matching this descriptor's ascending order.

⚠️ No CI run has gone green on tinyhttps#14. Its pull-request run was cancelled without ever being allocated a runner, and master's run is queued behind the same shortage. The evidence for the change is local: 17 tests from 6 test suites plus 3 in test_resolver, on the host and again on x86_64-linux-musl over openkal.

[Sunrisepeak]