1 files changed, 8 insertions, 3 deletions
diff --git a/tools/rtt.zig b/tools/rtt.zig
index 6ab0d87..d971c74 100644
--- a/tools/rtt.zig
+++ b/tools/rtt.zig
@@ -62,9 +62,14 @@ pub fn roundTrip(
             if (now - t0 > timeout_us) return null;
         } else if (now - last > quiet_us) break;
 
-        // Poll in millisecond units because that is what poll(2) takes; the TIMING above is
-        // microseconds and independent of this granularity.
-        const n = try port.readTimeout(&buf, 1);
+        // NON-BLOCKING, and that matters. `poll(2)` takes a timeout in MILLISECONDS, so waiting
+        // even 1 ms quantises this measurement into 1 ms buckets: a response that arrived just after
+        // a poll returned empty is reported up to a millisecond late. Against a round trip of a few
+        // milliseconds that is not a rounding error, it is a large fraction of the answer - measured
+        // as ~0.75 ms of unexplained gap between this figure and the sum of the board's own stage
+        // timings. A spin costs a busy host CPU for a few milliseconds per sample, which is the
+        // right trade for an instrument.
+        const n = try port.readTimeout(&buf, 0);
         if (n == 0) continue;
         if (first < 0) first = nowUs(port);
         bytes += n;