MozillaQualityAssurance:Environments: Difference between revisions

From MozillaWiki
Jump to navigation Jump to search
No edit summary
Line 298: Line 298:
RSS feeds and Microsummaries
RSS feeds and Microsummaries


===Next Steps/Issues===
==Next Steps/Issues==


*Solicit feedback from QA team and others as to what data makes sense. We need help determining how to extract the relevant data from the browser.
*Solicit feedback from QA team and others as to what data makes sense. We need help determining how to extract the relevant data from the browser.
Line 304: Line 304:
*Investigate an easy way to get addons popularity data. We have a good sense of what are the popular extensions, but conflicting data from sites such as extensionhunter.com make it difficult to assess-perhaps morgamic can help here.
*Investigate an easy way to get addons popularity data. We have a good sense of what are the popular extensions, but conflicting data from sites such as extensionhunter.com make it difficult to assess-perhaps morgamic can help here.
*Need some help building out the complex profiles (simulating things like 10K downloads). Who from QA/dev can assist here?
*Need some help building out the complex profiles (simulating things like 10K downloads). Who from QA/dev can assist here?
===VM Issues/Followup===
===VM Issues/Followup===
*Follow up needed to see where Justin is with VMware investigation
*Follow up needed to see where Justin is with VMware investigation

Revision as of 23:59, 21 June 2007

This page explores some of the topics around creating Firefox test environments that mimic "real world" and "typical" end user configurations.

Overall Goal

Configure the Mozilla testing systems as close as possible to a "typical" Firefox user in order to discover problems early and provide better configuration coverage.

Every user does not install Firefox into a fresh copy of Windows XP SP2 and people do not always upgrade to the latest and great MSFT patches. Their profiles may have been used for years and they have their web cache full of data or run a bunch of extensions. The goal of this exercise is to come up with a reasonable set of configs that best represent some typical configurations of our users.

Our current testing methodology uses a (random? chaotic?) combination of configuration environments that are not documented, controlled or reproducible.

Requirements

  • Scope out several "typical end user configurations" for Firefox.
  • Offer an easy way to do testing and throw away the environment
  • Keep the base environments up to date
  • Easily "apply a specific Firefox config" against the base environment
  • Some of the parameters to consider include:
    • Operating System (clean install vs. existing installation), version number, patches, etc... - Windows XP, Vista, Mac, Linux (which one?)
    • OS Languages (e.g. are they running US Windows XP or Japanese XP)
    • Firefox (clean new install), no profile versus Existing Profile
    • Profile character:
      • Clean web cache vs. 125 MB worth of stuff
      • 2 vs. 15 addons (which ones to select?)
      • Selection of saved passwords?
      • Other profile aspects?
    • One versus multiple OS users? (not sure)
    • One versus multiple Firefox profiles? (not sure)

Recommendation

  • Develop a "minimum set" (3-4 configs) and an "expanded set" (up to 10?)
  • Use our Vmware server and "snapshot capabilities" to define the following sets of VM's
  • Typical Profiles
    • No Profile (ie, new user to Firefox)
    • Profile A: 120 MB worth of cached web pages, 3 addons [typical]
    • Profile B: 200 MB worth of cached web pages, non-default home page, 10 addons [stress test]
  • Typical OS Setups
    • OS Setup 1: Windows XP SP2 + all high priority WindowsUpdate patches
    • OS Setup 2: Windows XP SP2 + no additional patches
    • OS Setup 3: Mac OS X 10.4.9 + all available SoftwareUpdate patches
    • Minimum Set:
      • Config 1: OS Setup 1 + no Firefox Profile
      • Config 2: OS Setup 1 + Firefox 2.0.0.x + Profile A
      • Config 3: OS Setup 1 + Firefox 2.0.0.x + Profile B
      • Config 4: OS Setup 3 + Firefox 2.0.0.x + Profile A
    • Expanded Set:
      • Config 5: OS Setup 3 + Firefox 2.0.0.x + Profile B
      • Config 6: OS Setup 3 + no Firefox Profile
      • Config 7: OS Setup 2 + no Firefox Profile
  • Create VM instances for OS Setup 1 sn
  • Use the VM instances in "read only mode" to test
  • VM Instance maintenance: Periodically, update the base config with the latest Microsoft Windows patches
  • Our Vmware Server allows us to store multiple copies of vm instances. So, perhaps we create several base instance (e.g. Windows XPSP2, Windows XP SP2+all current patches) and then script the creation of getting these base instances to the configs that we want (e.g. loading them with profile data, etc...).
  • For Mac based testing of Window Firefox, you can do the same thing but using Parallels. Create a clean Windows XPSP2 instance, save the config and copy it to a file server location. Using a copy of that instance, create Windows XPSP2+patches, save that instance. When testing Firefox, use a copy from the fileserver and then throw away the instance.
  • Keep the base instances updated with the latest patches from Microsoft periodically.
  • For Mac testing, not certain how we do that.

Notes from Profile and Configuration Meeting Held on 6-13-07

    • Attendees: timr, juanb, tchung, marcia, tomcat, galen

Typical OS Setup:

Tier 1:

  • Windows XP with latest updates
  • Windows Vista with latest updates
  • Mac OSX 10.4.x with latest updates
  • Linux - which distro? Needs more investigation
  • First implementation will focus on Tier 1 only
  • First implementation will focus on en-US only

Tier 2:

  • XP SP2 - No patches


Profile A Profile B Profile C Profile D
1. Page Cache 120 200
1. History
1. Cookies
1. Bookmarks 20 100 1000
1. Downloads 10 100 1000
1. Form Data
1. Passwords 5-10 20 100
1. Addons 5 25 50
1. Plugins (Java, Flash, etc) Flash, Java Flash, Java, Acrobat, Platform dependent media player All above plus Viewpoint
2. Tabs 2 1 5-10 25
2. Non Default Homepage
2. Installation History
2. Multi Users
3. Search Plugins 0 5 10
3. Preferences Does not change the prefs


Priority ranking of items 1Cache 1History 1Bookmarks 1Downloads 1Form Data 1Passwords 1Extensions and Themes 2Tabs 2Non Default Homepage 1Plugins (Java, Flash) 3Search Plugins 2Installation History - Clean Install/Paveover Install/Upgrade 2Multi-Users User 2 users Admin/Non Admin 2+Users 3Preferences

Profiles

  • A No Profile
  • B Typical Profile
  • C Large Profile
  • D Stress Profile

Configurations

  • 1 OS1 - FF 2.0.0.x - Profile A (Smoke)
  • 2. OS1 - FF 2.0.0.x - Profile B (BFT)
  • 3. OS1 - FF 2.0.0.x - Profile C (Smoke)
  • 4. OS2 - FF 2.0.0.x - Profile B (Smoke)
  • 5. OS3 - FF 2.0.0.x - Profile B
  • 6. OS4 - FF 2.0.0.x - Profile B

Other Issues

  • 110nLocales
  • Admin versus Non Admin
  • Priorities: Magnitude of problem (data loss)
  • Common Compatibility (Problems with Java)

Notes from Profile and Configuration Meeting Held on 6-21-07

attendees: marcia, tomcat and bc

Action Items from meeting

  • Tomcat: Talk to Shaver about Popular extensions on AMO
  • Review Asa blog post - Extensions you use
  • Follow up with Jay poll survey capability on QMO->marcia
  • Determine whether relevant data is stored in FF 2 and FF3->marcia
  • Profile D->Data from 15K page run->bc

Addons NoScript Adblock Flashblock

   * A No Profile
   * B Typical Profile -5-10
   * C Large Profile - 25
   * D Stress Profile -50

Cache Action

Page cache->50 is the Downloads-> Form data->formhistory.dat Sessionstore.js Sessonstore1.js Sessionstore.bak

Search History

Passwords - Certificates

What goes into the Installation Directory for Plugins On Linux, they are stored in a special place on their profile

Sort by Priority

Tabs

2 1 5-10 25

Passwords 5-10 Typical 20 100

Bookmarks

20 100 1000

Profile D->Data from 15K page run

Run Spider ->Set up the profile to automatically download to a particular location

10, 100, 1000

Search Plugin Typical-0 5 Stress-10

Flash Java Acrobat Viewpoint RealPlayer

Typical->Flash and Java C->Flash, Java, Acrobat, Media Player depending on platform, Java D->Every one including Viewpoint.

Need to ensure our Machine installation for Java-match what we end up in our installation


RSS feeds and Microsummaries

Next Steps/Issues

  • Solicit feedback from QA team and others as to what data makes sense. We need help determining how to extract the relevant data from the browser.
    • Recommendation made to put a survey up on QMO to gather data
  • Investigate an easy way to get addons popularity data. We have a good sense of what are the popular extensions, but conflicting data from sites such as extensionhunter.com make it difficult to assess-perhaps morgamic can help here.
  • Need some help building out the complex profiles (simulating things like 10K downloads). Who from QA/dev can assist here?

VM Issues/Followup

  • Follow up needed to see where Justin is with VMware investigation
  • Follow up needed for Mac VM Investigation->Someone should try out PearPC (http://pearpc.sourceforge.net/) and see how well it works. It lets you run Mac OS X on a PC

Target Completion Date for Phase 1 Implementation

  • Would like to see a set of 4 profiles created in time for testing of the Alpha6 release