From ccbf05e75868d335c6f23e711ee35eb9fa0b5d25 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Vin=C3=ADcius=20Ferr=C3=A3o?=
 <2031761+viniciusferrao@users.noreply.github.com>
Date: Sun, 6 Sep 2026 15:39:03 -0300
Subject: [PATCH] fix(xcat-dep): record a worker the kernel killed as failed

A child killed by a signal leaves 0 in the high byte of its wait status. The
per-codename worker loop read only that byte, so a cancelled or OOM-killed
codename was counted as built, and the run could reach validation with the
staging tree that worker never finished.

One helper decodes a wait status for both the worker loop and run_bounded, and
reports 128 plus the signal for a child the kernel killed.
---
 lib/XCAT/BuildUtils.pm | 13 +++++++++++--
 sbuild-all.pl          |  5 ++++-
 2 files changed, 15 insertions(+), 3 deletions(-)

diff --git a/lib/XCAT/BuildUtils.pm b/lib/XCAT/BuildUtils.pm
index c0b4bc9..fc5172d 100644
--- a/lib/XCAT/BuildUtils.pm
+++ b/lib/XCAT/BuildUtils.pm
@@ -22,6 +22,7 @@ our @EXPORT_OK = qw(
   every_step_failed
   forward_signals_to_workers
   block_handled_signals
+  exit_status
   restore_signal_mask
   hashes_equal
   print_step
@@ -202,6 +203,15 @@ sub stall_report {
     return $total;
 }
 
+# exit_status($status): the exit code of a waited-for child, or 128 plus the signal that killed it.
+# A child killed by a signal has 0 in the high byte, so reading only that byte reports a build the
+# kernel terminated as a build that succeeded.
+sub exit_status {
+    my ($status) = @_;
+    return 0 unless defined $status;
+    return ($status & 127) ? 128 + ($status & 127) : $status >> 8;
+}
+
 # block_handled_signals(): block INT, TERM and HUP and return the previous mask, for the window
 # between forking a child and being able to signal it. A cancellation arriving in that window would
 # otherwise kill the parent under a handler that does not know the child yet, and the child would
@@ -323,8 +333,7 @@ sub run_bounded {
         return { ec => 124, timed_out => 1, elapsed => time - $t0 };
     }
 
-    my $ec = ($status & 127) ? 128 + ($status & 127) : $status >> 8;
-    return { ec => $ec, timed_out => 0, elapsed => time - $t0 };
+    return { ec => exit_status($status), timed_out => 0, elapsed => time - $t0 };
 }
 
 sub display_quote {
diff --git a/sbuild-all.pl b/sbuild-all.pl
index 38eca37..5c9d9bd 100755
--- a/sbuild-all.pl
+++ b/sbuild-all.pl
@@ -662,7 +662,10 @@ sub build_deps {
         }
         my $pid = wait();
         if ($pid > 0) {
-            my $ec = $? >> 8;
+            # A worker the kernel killed leaves 0 in the high byte, so the shifted status alone
+            # would record a cancelled or OOM-killed codename as built.
+            require XCAT::BuildUtils;
+            my $ec = XCAT::BuildUtils::exit_status($?);
             my $cn = delete $pid2cn{$pid} // '?';
             $fail{$cn} = $ec if $ec != 0;
             $running--;