I've ran into this failure today:
140 0s ... - output mismatch (see 140.out.bad) --- /home/thuth/devel/qemu/tests/qemu-iotests/140.out 2019-05-07 17:57:08.000000000 +0200 +++ /home/thuth/tmp/qemu-build/tests/qemu-iotests/140.out.bad 2019-05-08 07:19:23.000000000 +0200 @@ -10,6 +10,8 @@ {"return": {}} qemu-io: can't open device nbd+unix:///drv?socket=TEST_DIR/nbd: Requested export not available server reported: export 'drv' not present +mkfifo: cannot create fifo '/home/thuth/tmp/qemu-build/tests/qemu-iotests/scratch/qmp-out-27628_0': File exists +mkfifo: cannot create fifo '/home/thuth/tmp/qemu-build/tests/qemu-iotests/scratch/qmp-in-27628_0': File exists {"return": {}} {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "SHUTDOWN", "data": {"guest": false, "reason": "host-qmp-quit"}} *** done ... and indeed, there were lots of stale qmp-in-* and qmp-out-* files in my scratch directory, ultimately causing the above failure after a couple of days. After some more testing, it seems like test 192 is not cleaning up correctly: $ ls scratch/ $ ./check -qcow2 192 [...] 192 0s ... Passed all 1 tests $ ls scratch/ nbd qemu-0.pid qmp-in-8772_0 qmp-out-8772_0 Any ideas how to fix this in a clean way (e.g. simply add a "rm -f scratch/..." at the end of the script) ? Thomas