summaryrefslogtreecommitdiff
path: root/docs/superpowers
diff options
context:
space:
mode:
Diffstat (limited to 'docs/superpowers')
-rw-r--r--docs/superpowers/followup/pr3851-audiod-short-fifo-read.md4
1 files changed, 2 insertions, 2 deletions
diff --git a/docs/superpowers/followup/pr3851-audiod-short-fifo-read.md b/docs/superpowers/followup/pr3851-audiod-short-fifo-read.md
index a9bb86c42..cf7fa7ace 100644
--- a/docs/superpowers/followup/pr3851-audiod-short-fifo-read.md
+++ b/docs/superpowers/followup/pr3851-audiod-short-fifo-read.md
@@ -12,8 +12,8 @@ audio_test_freertos: "Audio mismatch at sample 8702: expected 8702, got 8696").
previous packet's data — the host hears a replayed fragment, offset by exactly
one packet (6 samples at HS 48 kHz mono 16-bit: "expected 8702, got 8696").
- Reproduced whenever the producer lets the FIFO underrun; the example was fixed
- to hold the FIFO at threshold (PR #3851, `20ff2b498`), but any application
- with >~2 ms producer starvation can still hit the driver bug.
+ to hold the FIFO at threshold (this PR, #3866), but any application with
+ >~2 ms producer starvation can still hit the driver bug.
- A naive short-write is not a fix: sending fewer bytes than the flow-control
size is legal ISO behavior, but a 1-byte pad would byte-misalign the stream —
the packet must shrink to the bytes actually read (sample-aligned).