Extras-testing/QA checklist

The software hosted in extras-testing is not ready for normal users!
PLEASE use it only for testing purposes. Be ready to file proper bug reports instead of posting complaints.

Potential problems: crashes, battery drain, poor system performance, full disk space & more - SERIOUSLY!

Backing up your data is recommended. In case of trouble you might need to re-flash your device.

There are several elements to be considered by developers before promoting software to extras-testing. The same criteria are useful for beta testers before approving applications to go to Extras. Please take your time looking at them. Don't evaluate an application lightly! If you are busy or in a hurry just let others do the job.

Please check the points of this list. Don't rush to give thumbs up to an application before it is tested. But don't try to delay the release of an app with minor bugs or enhancement proposals separate from this discrete QA checklist.


[edit] QA Improvements

There is a ongoing QA (Quality Assurance) improvements discussion in talk.maemo.org (see here and here) and the mailing list. One task is to establish a testing squad. The results can be found at this wiki page: QA improvements

[edit] QA checklist

Use this checklist until the extras-testing approval interface is fixed. It helps to see what you have tested and what other testers should still test. This checklist condenses the blockers explained below, read the explanation below before you mark a test as done.

1. [ ] Bug database exist.
2. [ ] Licensing ok.
3. [ ] No illegal/dubious content.
4. [ ] Working provided features.
5. [ ] No missing announced features.
6. [ ] Optification ok.
7. [ ] No performance problems.
8. [ ] No power management issues.
9. [ ] No known security risks.

Additional comments:

Note: Only paste this template if there is some problem with the application or if you have not tested some of the points above, let's not spam the developers.

  • Copy&paste this to the comment box.
  • Put [x] for those tests you have done, elaborate on separate row if the test is FAIL.
  • Vote up if there were no FAILs. If there was even one FAIL, vote down.
  • UI usability issues cannot be used as a reason for vote down.
  • Always test functionality - that is, run the program and try if it works as it should.

imaginary example:

1. [x] Bug database exist.
2. [ ] Licensing ok.
3. [ ] No illegal/dubious content.
4. [x] Working provided features.
 FAIL: There is choice between tabs and spaces as separators but spaces are always used (see bug: http://url/123). 
 FAIL: When exporting file the program crashes (see bug: http://url/456)
5. [ ] No missing announced features.
6. [x] Optification ok.
 FAIL: the package uses 1512kb from root.
7. [ ] No performance problems.
8. [ ] No power management issues.
9. [ ] No known security risks.

Additional comments: 
I liked the program, even though, as is, I have to vote it down due to the bugs. I added few usability enhancements to bugzilla, see http://url/567 and http://url/678

[edit] Blockers

A "blocker" is any item not met in the QA checklist. If any one of the criteria in the QA checklist is not met, the application cannot go into Extras. A Critical bug at http://bugs.maemo.org/ against the application is required to justify your thumbs down.

  1. Lack of bug reporting database
    It needs to be identified in the http://maemo.org/packages/ page.
    http://bugs.maemo.org/ is the preferred option.
    A missing bugtracker is NOT a blocker as of 01.09.2010, but use of bugtracking facilities IS strongly recommended.
    Developers: If the bugtracker field is not specified, the address is defaulted to be that of the package itself on maemo.org: http://maemo.org/packages/view/<package>(/<version>) Resolutions for bugtracking are found at Packaging#Bugtracker_location
  2. Licensing and legal issues
    Evident licensing problems or copyright violation. The open source licenses need to be compliant. It needs to be clear that the product is not officially supported by Nokia, Maemo or other commercial entities and trademarks.
  3. Illegal or dubious content
    To avoid any doubt, the content must not depict explicit sexual activity; depict or endorse acts that cause or are intended to cause excessive pain or suffering; promote or endorse the misuse of alcohol, tobacco, illegal drugs or other addictive substances; promote intolerance or discrimination based on racial, political, ethnic, religious, gender differences or sexuality; promote invasion of rights of privacy; promote gambling or promote illegal activity.
  4. Working features, broken functionality or reproducible crashes
    There is some functionality that does not work or produces a crash. Report the problem in the application's bug reporting database. There you can provide a list of steps that lead to the problem.
  5. Missing announced features
    Core features advertised through UI or product page don't work or are missing.
  6. Optification
    Test using the df command:
    In Xterminal, run df -h | grep -e rootfs -e mmcblk0p2 before and after you install the package being tested, and see how much the usage of each partition changes. If more than, say, 300 kb (0.3 MB) are added to usage of rootfs, it's too much and the package needs to be optified.
    Test using the dpkg command:
    After installation of the package being tested, run in Xterminal dpkg -L packagename It will show all the files installed by the package. Inspect the result to confirm that some of the files are placed in the /opt or MyDocs directories.
    Please add those packages which fail optification to the list at Opt_Problem/Non-Optified_packages.
    Developers: An application is correctly "optified" if all files created during the normal operation of the application are created and stored in /opt or MyDocs, since these directories are not stored on the root filesystem. If the application or its dependencies ignore the recommendation to use the eMMC to install as much files as possible, filling the root partition with 500kb or more, this can be considered a blocker. Also see Opt Problem.
  7. System performance compromised
    Running the application visibly affects the performance and responsiveness of the system, either through specific actions or leaving the application open/running during a long period of time.
  8. Power management issues
    Battery life is diminished beyond acceptable standards by having the application running in the background or just through normal use. The use case of a full day without charging should not be compromised.
    If specific applications are explicitly using a lot of power to boost speed or performance (for instance games, Wi-Fi heavy apps) then they must warn the user explicitly with a first boot banner.
  9. Security risks
    The main security risks are financial damage, access to private data and harm to device components. If you find such risk in an application then you need to report it and the app can't be uploaded to Extras until a deeper analysis has been done with favourable results.

[edit] Non-blockers

The extras-testing QA process should not be used to delay the release of new applications and updates with minor/normal bugs or enhancement requests. These can be filed as bugs just as usual.

  • Developers are encouraged to adopt the Maemo 5 UI guidelines and optimize their application to finger use. However, lack of Maemo 5 UI optimization is not a reason to block an application in its way to Extras unless it is unusable even with stylus.
  • Maemo 5 offers stock icons covering most regular use cases, but developers can use the icons they prefer as long as they respect copyrights. Broken icons can be a major bug stopping a release to Extras but discussions about beauty/ugliness of a UI are out of scope in the QA process.

[edit] Warnings

If one of these warning criteria are not met, the developers will be asked to fix them urgently. Still, the app can make it slowly to Extras unless a blocker is raised.

  • Details at http://maemo.org/packages/ are missing or incomplete: summary, URL to project, updates info.
  • Application icon missing or not visible in the device. This is only optional for command-line applications.

Testers with good technical knowledge are invited to look deeper in the Maemo Quality considerations (((waiting for Fremantle update))) in order to find weak points in applications waiting to be promoted.

[edit] Command Line applications

Command line applications are a special case, so two extra checks should be performed, these are blockers:

  • The application uses the CLI icon in the application manager.
  • The application description clearly says that the application only runs from the command line.

[edit] Testing software

[edit] Power use

The N900 has extensive powersaving technologies. It can achieve in some circumstances very good battery life. For example, with a wifi AP that supports powersaving properly, connected over ssh and to a GSM network, battery life can be around 5 days idle.

Leaving an application inactive (accidentally or not) is not uncommon, and the application must not use significant power when doing nothing with the screen locked.

This is a list of some applications that may be of use to ensure that the application you are testing does not damage suspend time.

(for best results, run with screen locked, over ssh, with nothing plugged into USB, as this minimises normal system activity)

C#      | Ratio  | Avg/dura | Frequency | Ratio
     C0 |  57.8% |          |   600 MHz |   0.0% |
     C1 |  22.4% |    0.9ms |   550 MHz |   0.0% |
     C2 |  19.8% |    3.3ms |   500 MHz | 100.0% |
     C3 |   0.0% |          |   250 MHz |   0.0% |
     C4 |   0.0% |          | 

This is an example of an app that when idle does not powersave at all. The CPU is continually active.

The 'C0' state is the state where the CPU is actively running - 60% of the time, and it is spending 100% of its time when running at 500 MHz.

The battery may last 10-12 hours.

C1 through C4 are various powersaving states - with C4 being the deepest, it is spending at most 3 ms - 1/300th of a second in shallow powersaving.

In comparison - the same phone when idle - the stats look like:

C#      | Ratio  | Avg/dura | Frequency | Ratio
     C0 |   1.0% |          |   600 MHz |   0.0% |
     C1 |   0.0% |    0.2ms |   550 MHz |   0.0% |
     C2 |   3.0% |    5.9ms |   500 MHz |   0.0% |
     C3 |  17.6% |  252.0ms |   250 MHz | 100.0% |
     C4 |  78.3% |  810.5ms | 

The phone spends most of its time in the deepest powersaving state, is only actually running 1% of the time, and when it does, is running at 250 MHz.

The battery in this case will last over 110 hours.

[edit] top

Top or htop in extras can also be very handy. If when idle, the package you're testing is using 50% CPU - it is never going to be very good for battery life!

Cumulative time can be of use here - if every time you have looked, it has been using .4% CPU, but over half an hour it has used 10 min of CPU, something is wrong.