Jens-G opened a new pull request, #3882:
URL: https://github.com/apache/thrift/pull/3882
`transport::socket::tests::split_halves_must_not_clobber_each_others_timeout`
(from THRIFT-5954) sets a read timeout of 250 ms and asserts that it reads back
unchanged. Linux keeps socket timeouts in jiffies and rounds up, so on a kernel
built with `CONFIG_HZ=250` — the Debian default — the value comes back as 252
ms:
```
assertion `left == right` failed
left: Some(252ms)
right: Some(250ms)
```
## The change
Use **240 ms** instead of 250 ms. A multiple of 20 ms is a whole number of
jiffies at HZ 100, 250 and 1000 alike, so it is exact on every common kernel
configuration. The test's other two values, 80 ms and 500 ms, already qualify —
only the one value was wrong. A comment now records why the numbers are what
they are, so the next reader does not round them back to something tidier.
The assertion being tested is unaffected: it is about one half of a split
socket not clobbering the other's timeout, not about any particular duration.
## Verification
Measured on Debian 13, kernel 6.12, `CONFIG_HZ=250`, Rust 1.85.1:
| | unpatched | patched |
|---|---|---|
| `split_halves_must_not_clobber_each_others_timeout` | **FAILED** — `left:
Some(252ms)`, `right: Some(250ms)` | ok |
| `cargo test --lib` | — | 178 passed, 0 failed |
| `cargo fmt --check` | — | clean |
| `cargo clippy --lib -- -D warnings` | — | clean |
CI is green on this test today, so its runners keep 250 ms exact; this makes
the test independent of that.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
🤖 Generated with [Claude Code](https://claude.com/claude-code)
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
