As part of looking at how we ensure that Ubuntu Certification delivers a great experience to consumers, we are revising our coverage of Suspend/Resume and boot testing. This is a proposal at the moment , so please feel free to comment on the suggested tests or additional ones that you would fine worth including.
Here is the list of proposed tests that a certify system would need to pass for 11.10 with respect suspend and resume, this include existing and new tests:
| Test | Description |
| Suspend_once | Main sleep test, all “after” tests depend on this one. Triggered manually, auto-wakes after 30-60 seconds (not all systems support automatic trigger and wake up) |
| hibernate_once | Triggered manually, auto-wakes after 5 minutes |
| CPUs Checking before/after suspend | Check in /proc/interrupt before and after suspend how many cores are online. Fail ‘after’ test if different. |
| Memory before/after suspend | Check in /proc/meminfo that all system memory is still available. |
| USB Before and After suspend | One per port, write to USB storage device (thumb drive) and verify. |
| Display | Ensure that the display is working after resume (N/A for servers) |
| resolution_*_suspend | ensure that resolution is same before/after suspend (N/A for servers) |
| cycle_resolutions_after_suspend | Can cycle through resolutions after a suspend (N/A for servers) |
| *wireless_*_suspend | Checks that wireless network is still available and can connect to it |
| network_*_suspend | Checks that wired network is still available and can connect to it |
| Wake-on-lan | put system to sleep then send an IP packet from a remote system to wake it (only applicable to servers) |
| audio_*_suspend | Audio device still works after detected (N/A for servers) |
| record_playback_after_suspend | records and playback after suspend (N/A for servers) |
| bluetooth_obex_*_suspend | Bluetooth obex object can be sent to another device after suspend (N/A for servers) |
| stress/suspend | Does 30 S3/Resume cycles,can only fail once |
| 30 Soft Reboots | Restart during testing 30 times |
Victor Tuson Palau 
stress/suspend Does 30 S3/Resume cycles,can only fail once
why is it allowed to fail suspending?
it is allow to fail 1 out of 30, basically that is the rate that we think is acceptable. Some time race conditions might happen that will mean that is the case.
Hibernate isnt really a massive priority and there was talk about removing it from the interface since its causing so much problems on most machines. Id say that should be removed from the list.
You are correct, we will test hibernate, and let people know if it failed. But hibernate faliure won’t block certification
I realize these are proposed tests. Are the final agreed on tests going to be archived on the public wiki for later reference? Are the current list of certification tests available on the public wiki?
-jef
I would like to have the final agreed test on a public place for reference. Might not be the wiki though
Are the current tests in a publicly archived place for reference?
-jef
Even with light use, one failure in 30 suspend cycles will mean “once a month” in user’s hands. People who suspend more than once a day, will see accordingly higher failure rates.
Hi David,
Not really, this is 30 cycles without powering off. So if you run your computer for a month without reboot and you suspend once a day, that will be true.
I take the feedback, we might remove the 1 faliure, but that only means that might fail on day 31 😉
For systems that have a hardware wifi switch (airplane mode) you might want to test that resuming with wifi switched off and then enabling wifi will result in working wifi. This is a class of problem that is not rare.
For systems with just a soft wifi swtich (Fn key) testing that the switch actually disables wifi/bluetooth and then re-enables it would be useful.
we consider this but we have not include it at the moment, as it add the level of manual testing required. We have kept it on the “next round” to be consider for “P” release