From 03c780fca8bb374d852172ff79d6427c7b2fa078 Mon Sep 17 00:00:00 2001
From: samuelburnham <45365069+samuelburnham@users.noreply.github.com>
Date: Thu, 1 Oct 2026 14:55:59 -0400
Subject: [PATCH] fix(bencher-track): Read the whole multi-line thresholds
 input

The `thresholds` input spans several lines, one per measure. Replacing
the unquoted array expansion with `read -a` kept only the first line,
so every measure after the first lost its flags and the first lost its
boundaries; bencher rejected the model with "no boundary provided" and
Benchmark main failed at the tracking step.

`read -d ''` consumes the input to its end and still splits on all
whitespace, reproducing the original token list without globbing. It
reports end of input as a failure, hence the trailing `|| true`.
---
 .github/actions/bencher-track/action.yml | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/.github/actions/bencher-track/action.yml b/.github/actions/bencher-track/action.yml
index 45be01838..4d645f6e5 100644
--- a/.github/actions/bencher-track/action.yml
+++ b/.github/actions/bencher-track/action.yml
@@ -72,8 +72,11 @@ runs:
         # No data points since the reset → omit thresholds so the reset commit
         # lands without an alert (--thresholds-reset clears stale models); they
         # re-arm next run. Otherwise substitute the window into each measure.
+        # The input spans several lines, so read to end of input (`-d ''`)
+        # rather than to the first newline; `read` then reports end of input
+        # as a failure, hence `|| true`.
         thresholds=()
-        [ "$sample" -gt 0 ] && read -r -a thresholds <<< "${THRESHOLDS//__WINDOW__/$sample}"
+        [ "$sample" -gt 0 ] && read -r -d '' -a thresholds <<< "${THRESHOLDS//__WINDOW__/$sample}" || true
 
         bencher run \
           --project ix \