On Sun, Apr 10, 2016 at 08:58:10PM -0400, Phillip Susi wrote: > On 03/01/2016 04:45 PM, Brian C. Lane wrote: > > Sometimes the device will vanish after the wait_for_dev_to_appear > > exits. Add udevadm settle in an attempt to make sure the udev > > system is done flapping around and the device will stay in place. > > > > Related: rhbz#1260664 --- tests/t-lib-helpers.sh | 2 ++ > > tests/t-local.sh | 2 ++ tests/t6001-psep.sh > > | 2 ++ tests/t6004-dm-many-partitions.sh | 1 + > > tests/t6005-dm-uuid.sh | 2 ++ > > tests/t6100-mdraid-partitions.sh | 1 + 6 files changed, 10 > > insertions(+) > > Why add the require_udevadm_settle to these 4 specific tests? Other > tests call wait_for_dev_to_appear. Don't they need it to? Shouldn't > the require be added to t-lib-helpers or somesuch to require it for > all tests that call wait_for_dev_to_appear?
The other users already pull it in via require_scsi_debug_module_, although it looks like I missed t6004 somehow, it needs it as well. There's no way to add a generic check to t-lib-helpers, since we want to skip tests early if it isn't available. Although the other option is to made the settle a library call, and if udevadm isn't available have it be a noop instead of skip. I'm not sure if there are any distributions using parted that don't use udev though, so that might be completely pointless. -- Brian C. Lane | Anaconda Team | IRC: bcl #anaconda | Port Orchard, WA (PST8PDT)
