626059 - add classID property to package.json to support XUL (non-bootstrapped) extensions in Mozilla2, use the same logic for generating JID, harness GUID, and harness contract ID in 'cfx run' and 'cfx test' when an id is present in package.json.

Status: RESOLVED FIXED

(Add-on SDK Graveyard :: General, defect)

Description

[Nickolay_Ponomarev]

With the switch to chrome.manifest-based registration of XPCOM components in Mozilla 2, the dynamic generation of classIDs <https://github.com/mozilla/addon-sdk/blame/master/python-lib/cuddlefish/__init__.py#LID534> for the harness component became problematic for non-bootstrapped (XUL) extensions: if your bootstrap.js isn't invoked, you have to register the harness component via chrome.manifest, which requires hardcoding a classID there.

As a workaround for this problem I added support for hardcoding the CID in package.json.

While doing that I elmiminated the unnecessary difference between 'cfx test' and 'cfx run' WRT harness_guid, jid, and harness contractID they use. As I understood, the idea was to allow 'cfx test' to work without an id in package.json. I made it so that the hardcoded GUID/harness ID is only used when the id isn't present in package.json.

The packages/development-mode/lib/bootstrap.js change is fixing XPCOM unregistration to work in the case of randomly generated GUID for each run: it should unregister using the classID of the previous run, not the classID of the current run.

The examples/reading-data/tests/test-main.js change is needed because the tests now use the JID (and the resource:// prefix is generated from the JID) from package.json, when its provided, as is the case with the reading-data package.

Comment 1

[Nickolay_Ponomarev]

Attachment: pointer to the pull request

https://github.com/mozilla/addon-sdk/pull/92

Comment 2

[Brian Warner [:warner :bwarner]]

Comment on attachment 504124 [details]
pointer to the pull request

looks good to me! It also changes the behavior of 'cfx test' (when used on a module that *does* have an ID in its package.json) to use the same JID as 'cfx run' or 'cfx xpi', which is probably a good thing.

Comment 3

[Nickolay_Ponomarev]

> It also changes the behavior of 'cfx test'

Yes, I mentioned that in comment 0 and it's intentional. Before I had to check two contract IDs when trying to get the harness service from a XUL extension. Tests also passed with this change.

Thanks for the review! I assume this has to wait until the thaw to get landed?

Comment 4

[Nickolay_Ponomarev]

Brian, now when the tree is open could you land it for me? I updated the pull request today and it doesn't fail any new tests (compared to the addon-sdk tip).

Comment 5

[Brian Warner [:warner :bwarner]]

landed, in https://github.com/mozilla/addon-sdk/commit/d1dc326f420c9c1a897cfe94531532a47e7d6b1b

Comment 6

[Nickolay_Ponomarev]

Thanks, Brian!