syz-manager: silently update the corpus version

The v4->v5 transition has inadvertetly led to impossibly slow corpus
triage attempts on syzkaller instances that use emulated VMs. They never
finish the corpus triage, so they never bump the version.

Now that the original change has already reached its purpose, let's no
longer force the re-smashing of all corpus items.

1 files changed, 7 insertions, 1 deletions

diff --git a/syz-manager/manager.go b/syz-manager/manager.go
index af9076fe6..037b3a305 100644
--- a/syz-manager/manager.go
+++ b/syz-manager/manager.go
@@ -669,7 +669,13 @@ func (mgr *Manager) preloadCorpus() {
 		fallthrough
 	case 4:
 		// Version 4->5: fix for comparison argument sign extension.
-		corpusFlags &= ^fuzzer.ProgSmashed
+		// Introduced in 1ba0279d74a35e96e81de87073212d2b20256e8f.
+
+		// Update (July 2024):
+		// We used to reset the fuzzer.ProgSmashed flag here, but it has led to
+		// perpetual corpus retriage on slow syzkaller instances. By now, all faster
+		// instances must have already bumped their corpus versions, so let's just
+		// increase the version to let all others go past the corpus triage stage.
 		fallthrough
 	case currentDBVersion:
 	}