[monsterlag]

Note

AI-generated, not reviewed by a human. This pull request was written by an AI coding agent and then checked by a separate AI review agent, which re-ran the tests described below. No human has reviewed the code. It is submitted by @monsterlag, who runs these agents. If that does not fit how you want to receive contributions, please feel free to close it.

What

The README's Tips section only said that --latency-correction avoids the "Coordinated Omission Problem", without saying what the flag changes. #211 asks what it does, and a comment there asked for a short README explanation, which the maintainer agreed would be good.

This adds a short paragraph under the existing --latency-correction tip:

• with -q or --burst-delay, requests are scheduled at a fixed rate but can only be sent when one of the -c connections is free;
• by default latency is measured from when a request is actually sent, so time spent waiting for a free connection is left out;
• with --latency-correction, latency is measured from the scheduled time, so that waiting is included;
• it does nothing unless -q or --burst-delay is set.

It also corrects the flag's help text. It said "It's ignored if -q is not set", but the flag also applies with --burst-delay. It now says "It's ignored if neither -q nor --burst-delay is set." The copies of the help output in README.md and README_zh.md are updated to match.

This follows the code: RequestResult::duration() uses start_latency_correction (the scheduled time from the rate-limit queue) when it is set, and the corrected code paths are only used when a query limit (-q or --burst-delay) is set.

How tested

To check the text is accurate, I ran oha against a local Python HTTP server that waits 1 s before each response:

Command	Average	Slowest
oha --no-tui -n 6 -c 1 -q 4 URL	1.0343 s	1.0410 s
oha --no-tui -n 6 -c 1 -q 4 --latency-correction URL	2.9842 s	4.9770 s
oha --no-tui -n 6 -c 1 --burst-delay 200ms --burst-rate 3 URL	1.0361 s	1.0447 s
oha --no-tui -n 6 -c 1 --burst-delay 200ms --burst-rate 3 --latency-correction URL	3.5076 s	6.0140 s
oha --no-tui -n 3 -c 1 --latency-correction URL (no -q)	1.0286 s	1.0420 s

Without the flag every request reports about 1 s. With it, the reported latency includes the time each request waited for the single connection. Without -q or --burst-delay the flag has no effect.

After changing the help string, I ran oha --help on a debug build. It shows the new text, and that line matches the help blocks in README.md and README_zh.md. cargo fmt --check passes.

I did not add the new explanation to README_zh.md. Its Chinese one-line tip could get a translation in a follow-up.

Partially addresses #211

[hatoo]

nice!