connect_wireless can unintentionally choose a non-wireless connection to connect to

Based on discussion with the original reporter, it seems "eth0" was mentioned by the wireless_before_suspend test, this leads me to believe that at the time of running the test, eth0 was connected and maybe the wireless interface wasn't. Another possibility was that some other wireless connections were configured thus confusing the test script.

Even though strictly speaking it's unconfirmed I'll set to importance: High as wireless testing needs to be reliable.

[danne@danne-fnuttbook] ~/Downloads> nmcli con
NAME UUID TYPE TIMESTAMP-REAL
Auto Dannes Nalle 232a0551-6baf-400f-ae87-016f1fddb87c 802-11-wireless Fri 23 Sep 2011 08:07:26 AM CEST
Auto B2_private_8F 3f19c700-3591-4ee5-b6f1-c628ce09a47d 802-11-wireless never
Auto wiberg b0e56574-30e7-4218-95ae-4d111cb34878 802-11-wireless Sun 11 Sep 2011 11:01:53 AM CEST
Wired connection 1 ea3cb0b8-4028-4c0d-b046-b4c8977f7483 802-3-ethernet Tue 18 Oct 2011 07:14:03 PM CEST
wiberg.nu 971ebaeb-9ce2-40c3-bc43-be5f621c33ff vpn never
Auto wiberg.nu 00d990e8-dac8-4e7d-b74a-aae85dfea53a 802-11-wireless Tue 18 Oct 2011 11:36:01 PM CEST

I tried removing all but the active wireless connection, but it didn't help. But I noticed that the test before and after suspend seems to disconnect the wireless, so I think the problem is that the test script disconnects the wireless and then reports it as not functioning. I will do the same tests on my work laptop tomorrow to compare the results.

Hi Daniel,

The logic of the test is as follows:

If checks if you *have* a wireless connection at the point of running the test with 'nmcli -f SSID,ACTIVE dev wifi list'

For you this should be 'wiberg.nu'

It then uses this to get the connection from the list of existing ones, and here's the problem - it picks up 'wiberg.nu 971ebaeb-9ce2-40c3-bc43-be5f621c33ff vpn never' as the connection to use in the test. It then knocks down all your existing connections and brings back up your 'wireless' connection (which in this case is not one).

Try:

> active_connection=$(nmcli -f SSID,ACTIVE dev wifi list | grep yes)
> ap=$(echo $active_connection | awk -F\' '{print $2}')
> conn=$(nmcli -t -f UUID,NAME con list | grep wireless | grep -e "$ap$" | awk -F\: '{print $1}')

versus:

> active_connection=$(nmcli -f SSID,ACTIVE dev wifi list | grep yes)
> ap=$(echo $active_connection | awk -F\' '{print $2}')
> conn=$(nmcli -t -f UUID,NAME con list | grep -e "$ap$" | awk -F\: '{print $1}')

If it's ok to change the 'wiberg.nu' connection name to something like 'wibergvpn' then this will let the test run. In the meanwhile please check the modified code and let me know if it works.

I'll test this as soon as I get home from work.

After a little modification your solution seams to work, type must also be included:
conn=$(nmcli -t -f UUID,TYPE,NAME con list | grep wireless | grep -e "$ap$" | awk -F\: '{print $1}')

I'll see if the result shows up on friendly.ubuntu.com soon.

I think this would be a good fix as it's not too uncommon that your company wireless and VPN have similar names.

The attachment "Patch to only look for wireless connection." of this bug report has been identified as being a patch. The ubuntu-reviewers team has been subscribed to the bug report so that they can review the patch. In the event that this is in fact not a patch you can resolve this situation by removing the tag 'patch' from the bug report and editing the attachment so that it is not flagged as a patch. Additionally, if you are member of the ubuntu-sponsors please also unsubscribe the team from this bug report.

[This is an automated message performed by a Launchpad user owned by Brian Murray. Please contact him regarding any issues with the action taken in this bug report.]

Results are up on friendly.ubuntu.com now, I think it works,

Hi Daniel,

Yes, thanks for catching the missing TYPE. We'll schedule this fix for an SRU and it will also be included in Precise.

This bug was fixed in the package checkbox - 0.13

---------------
checkbox (0.13) precise; urgency=low

  New upstream release (LP: #892268):

  [Marc Tardif]
  * Generate a submission.xml file that contains all device and attachment
  * Write the report before reporting the validation error.
  * Changed device.product to dmi.product for the formfactor (LP: #875312)

  [Daniel Manrique]
  * Use gettext for string (LP: #869267)
  * Move progress indicator to main checkbox dialog instead of a
    transient window (LP: #868995)
  * Ignore malformed dpkg entries in package_resource (LP: #794747)
  * Reset window title after finishing a manual test (LP: #874690)
  * Handle "@" in locale names (as in ca@valencia).

  [Jeff Lane]
  * Went through all the job files and:
    * Updated descriptions to match Unity UI structure
    * Added descriptions where necessary
    * Added further details to some descriptions
    * Moved some jobs to more appropriate files
    * Fixed job names in older job files to match new naming scheme
      (suite/testname)
    * Added jobs to local.txt to ensure all job files are now parsed
      (this allows easier addition of existing tests to whitelists)
    * Changed remaining manual job descriptions to match the new format
  * Updated CD and DVD write tests to be more clear about when to skip
    them (LP: #772794)

  [Ara Pulido]
  * Rewrote all job descriptions to match OEM QA syntax

  [Brendan Donegan]
  * Fix the code that assigns keys in checkbox-cli so that it never assigns
    keys which have other uses. (LP: #877467)
  * Show details of unmet job requirements (LP: #855852)
  * Ensure that connect_wireless chooses a wireless connection from the list
    of available connections (LP: #877752)
  * Have the bluetooth/detect tests require a device with the category
    BLUETOOTH to run, thus preventing the test from failing on systems with
    no Bluetooth device (LP: #862322)
  * Rename attachment jobs to not have a forward slash in their name
    (LP: #887964)
  * Guard against trying to write files to logical partitions on USB sticks
    (which will obviously fail) in usb_test (LP: #887049)
  * Make the OpenGL test ignore the return value of glxgears and improve
    the test description (LP: #890725)
  * Allow input/mouse test to run if a TOUCH device is present
    (LP: #886129)

  [ Javier Collado ]
  * Broken job dependencies fixed (LP: #888447)
  * Regex support when specifying blacklists and whitelists on the
    commandline (LP: #588647)
 -- Daniel Manrique <email address hidden> Thu, 18 Nov 2011 12:46:21 -0500

Hello Daniel, or anyone else affected,

Accepted checkbox into oneiric-proposed, the package will build now and be available in a few hours. Please test and give feedback here. See https://wiki.ubuntu.com/Testing/EnableProposed for documentation how to enable and use -proposed. Thank you in advance!

Hi, I'm Daniel, but not the original reporter :)

Anyway. I took a Dell Inspiron 1545 and connected to a wireless network named Ubuntu. I then renamed the "Wired Connection 1" default ethernet to "Ubuntu". nmcli con list shows two Ubuntu networks. The first one that appears on the list is the Ethernet one.

As per the reproduction case I unplug the Ethernet cable. I'm able to ping some hosts on the wireless network as I'm still connected on that one.

I then run /usr/share/checkbox/scripts/connect_wireless from Checkbox 0.12.8 and I get:

Error: Device 'eth0' (/org/freedesktop/NetworkManager/Devices/1) disconnecting failed: This device is not active
Device state: 30 (disconnected)
Error: Unknown connection: 343b92b1-1049-406c-bfd9-a636b93cacdf 8ddcecca-5513-495f-9525-b611bf620868.

The wireless connection remains disconnected and I can't ping anything.

I then installed Checkbox 0.12.9 from -proposed. I brought up the wireless connection. I then ran the same script and got the same result :( (notice the slight difference in the last line, here it can't even find a connection name)

Error: device 'eth0 (/org/freedesktop/NetworkManager/Devices/1) disconnecting failed: this device is not active
Device state: 30 (disconnected)
Error: Unknown connection: .

I went looking into the code and the problem is that, even though it's now filtering only the wireless connections (line 11 of connect_wireless), it had to add the TYPE field to nmcli output to be able to do so. However, later in the same line, it looks for the network name at the *end* of the line (grep -e '$ap$' where $ap contains the network name). Since the TYPE field was added at the end of the line, the script fails in a slightly different way.

The fix for the fix is to swap TYPE and NAME in line 11 so the field list looks like UUID,TYPE,NAME.

As the SRUed fix is not working as intended, I'll mark verification-failed. Also, the faulty code is also present in checkbox trunk, it will have to be fixed there as well.

Please let me know how to proceed with this bug report.

Thanks!

oneiric has seen the end of its life and is no longer receiving any updates. Marking the oneiric task for this ticket as "Won't Fix".