summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorMaxim Kuvyrkov <maxim.kuvyrkov@linaro.org>2021-07-20 07:54:41 +0000
committerMaxim Kuvyrkov <maxim.kuvyrkov@linaro.org>2021-07-20 09:02:52 +0000
commit7b4925acaa64ef9556529e33478fa8ec938bbe6a (patch)
treeaba520cb9e0c34499fa53551352d1551cf58687d
parent00b6a45c5d7534dcfab6df20370f12f071b16bf4 (diff)
round-robin-bisect.sh: Update comment
Change-Id: I8005b57a6aa1e073dfd83e259820a95b4afef37d
-rwxr-xr-xround-robin-bisect.sh7
1 files changed, 3 insertions, 4 deletions
diff --git a/round-robin-bisect.sh b/round-robin-bisect.sh
index e14895cd..f08e9a77 100755
--- a/round-robin-bisect.sh
+++ b/round-robin-bisect.sh
@@ -164,10 +164,9 @@ fi
# To catch the "2000" to "1000" regression someone would need to manually
# trigger bisect between rev_22 and rev_30.
#
-# TODO: We could skip revisions (exit 125) that are worse than metric
-# for $bad_rev (result metric <1000 in the above scenario), so we would
-# skip revisions between rev_20 and rev_22. This might cause other edge
-# cases to be handled sub-optimally, though.
+# We reduce the impact of the above by assigning negative values to result
+# metric for builds that look very broken, and do not bisect regressions into
+# the "negative" side.
cat > $artifacts/test.sh <<EOF
#!/bin/bash