Closed Bug 684867 Opened 13 years ago Closed 13 years ago

Add a test restartless extension to litmus-data

Categories

(Mozilla QA Graveyard :: Mozmill Tests, defect)

defect
Not set
normal

Tracking

(Not tracked)

VERIFIED FIXED

People

(Reporter: AlexLakatos, Unassigned)

References

Details

(Whiteboard: [mozmill-aom])

Attachments

(5 files)

Add a restartless extension to the litmus-data repo (mozqa.com/data)

*restartless.xpi
Restartless addon made using the online addon builder. Changed the maxVersion to * in install.rdf
Assignee: nobody → alex.lakatos
Status: NEW → ASSIGNED
Attachment #558486 - Flags: review?(anthony.s.hughes)
Attachment #558486 - Flags: review?(anthony.s.hughes) → review+
Comment on attachment 558486 [details] [diff] [review]
patch v1.0 [checked-in]

Landed:
http://hg.mozilla.org/qa/litmus-data/rev/09a280a9c78f
Attachment #558486 - Attachment description: patch v1.0 → patch v1.0 [checked-in]
Status: ASSIGNED → RESOLVED
Closed: 13 years ago
Resolution: --- → FIXED
Please mark this test VERIFIED if you are able to see and install the extension from the repository.
Marking as VERIFIED as I am able to see and install the extension from the repository.
Status: RESOLVED → VERIFIED
Please update the install.rdf and bump the minVersion to 4.0. We do not natively support restartless extensions in Firefox versions before 4.0.

Further I would like to be the reviewer of any extension patch in the future. As the former owner of the Add-ons Manager I have deeper insight in all the details. Thanks.
Status: VERIFIED → REOPENED
Resolution: FIXED → ---
Attached patch minVersion fixSplinter Review
bumped the minVersion to 4.0
Attachment #559439 - Flags: review?(hskupin)
Attachment #559439 - Flags: review?(hskupin) → review+
Follow-up landed as:
http://hg.mozilla.org/qa/litmus-data/rev/7c88ab3466a4
Status: REOPENED → RESOLVED
Closed: 13 years ago13 years ago
Resolution: --- → FIXED
As seen by current reports people don't find any ui for the extension to verify its functionality. Reason is that for existing profiles we cannot assume that the add-ons bar is visible. Therefore please remove the icon from the add-on bar and make it a context menu.

When updating the Litmus test please also reference the XPI directly and give details where people are able to verify its functionality. Thanks.
Status: RESOLVED → REOPENED
Resolution: FIXED → ---
Removed the icon and added a context menu entry "Open QMO"
Attachment #561480 - Flags: review?(hskupin)
(In reply to Alex Lakatos from comment #9)
> Created attachment 561480 [details] [diff] [review]
> patch v2.0
> 
> Removed the icon and added a context menu entry "Open QMO"

Alex, can you please give your patches meaningful names in the future? Instead of just calling the file "patch", "restartlessXPI.patch" would be better.

Thanks
Comment on attachment 561480 [details] [diff] [review]
patch v2.0 [checked-in]

Taking the review from Henrik. I've installed and tested the extension and it seems to work fine. Checking in.
Attachment #561480 - Flags: review?(hskupin) → review+
Comment on attachment 561480 [details] [diff] [review]
patch v2.0 [checked-in]

Landed:
http://hg.mozilla.org/qa/litmus-data/rev/25156151090e
Attachment #561480 - Attachment description: patch v2.0 → patch v2.0 [checked-in]
Status: REOPENED → RESOLVED
Closed: 13 years ago13 years ago
Resolution: --- → FIXED
Marking as VERIFIED as the extension installs from mozqa.com and works fine.

Edited the Litmus testcases that used it.
Status: RESOLVED → VERIFIED
(In reply to Alex Lakatos from comment #13)
> Edited the Litmus testcases that used it.

It's always helpful to have a reference here to at least one of those tests on litmus. Also don't forget to update the AOM spreadsheet too.
Also we have a couple of other tests on Litmus which should be updated to use this restartless extension.
I only edited the "[addons:extensions] Install a restartless extension" testcases. I'll search for others and update them as well.
The restartless extension is quite heavy in size. Per default it will include everything even if is some libs are not necessary. Alex, please check the following blog post and how to use the -strip-xpi option to reduce the size of the extension drastically. We shouldn't check-in that current version of the extension into our mozmill-tests repository.
Status: VERIFIED → REOPENED
Resolution: FIXED → ---
Mozilla/5.0 (Windows NT 6.1; rv:8.0) Gecko/20100101 Firefox/8.0
http://mozqa.com/data/firefox/addons/extensions/restartless.xpi

Installed the extension from the above location and at first it didn't show the "Open QMO" option in context menu. Then it started to show on some pages, but in others not. In the end, after a trying for a while it showed in all pages.

This issue is intermittent, I could not reproduce it on a new profile.
(In reply to Mihaela Velimiroviciu [QA] from comment #19)
> This issue is intermittent, I could not reproduce it on a new profile.
Mihaela, I'm rewriting the addon. We'll see if this still happens then, as I can't reproduce it now.
Attached patch patch v3.0Splinter Review
Remade the extension with Addon SDK 1.2.1 but it only stripped a few pixels
Attachment #569673 - Flags: review?(anthony.s.hughes)
Comment on attachment 569673 [details] [diff] [review]
patch v3.0

How have you build the XPI?
(In reply to Alex Lakatos from comment #21)
> Created attachment 569673 [details] [diff] [review] [diff] [details] [review]
> patch v3.0
> 
> Remade the extension with Addon SDK 1.2.1 but it only stripped a few pixels
it stripped only a few *kilobytes*, sorry for the confusion.

(In reply to Henrik Skupin (:whimboo) from comment #22)
> Comment on attachment 569673 [details] [diff] [review] [diff] [details] [review]
> patch v3.0
> 
> How have you build the XPI?
I used cfx --strip-xpi xpi
Comment on attachment 569673 [details] [diff] [review]
patch v3.0

Hm, ok. In that case lets leave it as we have it now. Looks good.
Attachment #569673 - Flags: review?(anthony.s.hughes) → review+
Landed as:
http://hg.mozilla.org/qa/litmus-data/rev/28e4ffb61a6b
Status: REOPENED → RESOLVED
Closed: 13 years ago13 years ago
Resolution: --- → FIXED
Right now the extensions points to http://quality.mozilla.org, but we get redirected to https. A security warning will display, which hangs our tests that use the restartless extension.

So the extension must point to https://quality.mozilla.org.
Status: RESOLVED → REOPENED
Resolution: FIXED → ---
Blocks: 680918
(In reply to Remus Pop (:RemusPop) from comment #26)
> A security warning will display, which hangs our tests that use the restartless extension.

Sorry to misinform you, the dialog is not from redirection to https. It is actually from viewing mixed content (secured + unsecured).
Status: REOPENED → RESOLVED
Closed: 13 years ago13 years ago
Resolution: --- → FIXED
Does that dialog not break any tests? If not, let's mark this VERIFIED. If so, we should add a boiler plate to this test to inform devs to account for it.
We have no landed test that uses this restartless extension. Although it would be blocking bug 680918, I made a workaround to skip the dialog by setting security.warn_viewing_mixed to false.
I still don't understand what's the problem here. When does such a security dialog comes up? Please be more specific.
This comes up the first time you visit https://quality.mozilla.org.
Ran this on a nightly using mozmill -b command.
Have you filed a bug against the QMO website? Regardless for now I would propose we change the link to use HTTP instead of HTTPS. Please file a follow-up bug.
As I said earlier, we use http in the extension but we get redirected to https anyway.
It doesn't matter how we get the page (http or https) because the results are the same.

There is already a bug filed for mixed content. I have added it to blocking.
Depends on: 701391
(In reply to Remus Pop (:RemusPop) from comment #33)
> It doesn't matter how we get the page (http or https) because the results
> are the same.

It does matter because it's an extra hop and a potential failure point for the test. Loading the desired test page (without redirect) is always the way we should go, except in cases when we want to test a redirect.

Thanks for filing the dependency.
This is because of the recent change of QMO to https, which has had some issues with a few of our plugins.

For the record, no mozmill tests should be using a live site for testing except for mozqa.com. Even though QA owns QMO, it can change without warning because no one running QMO (that's me) knows about your dependencies on it for day to day testing. The whole reason we set up mozqa.com and added ssl and such to it was to provide a fairly complete webserver for mozmill and other testing.
Al's point about mozqa.com vs QMO is well taken. Can you modify the extension to point to http://mozqa.com/data/firefox/layout/mozilla.html instead? Is there a unique purpose to pointing to QMO? If not, the above page should work fine for testing purposes.

If this seems reasonable, please reopen the bug. Thanks.
Lets do it in a new bug please.
Filled and patched bug 704831 for that
Dependency has been resolved and extension now points to mozqa.com. I think this can be marked verified.
Assignee: alex.lakatos → nobody
Status: RESOLVED → VERIFIED
Component: Infrastructure → Mozmill Tests
QA Contact: infrastructure → mozmill-tests
Whiteboard: [mozmill-aom]
Product: Mozilla QA → Mozilla QA Graveyard
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: