Write tests for privacy.thirdparty.isolate

added MikePerry201501R TorBrowserTeam201501R component::applications/tor browser owner::arthuredelstein priority::medium resolution::fixed status::closed tbb-4.5-alpha-3 type::task labels

Hm... I actually think we should try to get that into an xpcshell test as these are more lightweight and Mozilla recommends using mochitests only if we can't use xpcshell (see: http://ehsanakhgari.org/wp-content/uploads/talks/test-mozilla/ slide 15). And having different domains should be not a problem either as one can add identities to the HTTP server (see e.g.: https://mxr.mozilla.org/mozilla-central/source/security/manager/ssl/tests/unit/head_psm.js#434). If that means "You get your xpcshell test only if you write it!" then be it so. :) This might even help us directly as our xpcshell testing is not as broken as our mochitest testing IIRC (but I might be wrong here).

Trac:
Cc: N/A to gk, boklm

I had #13472 (moved) in mind when I wrote my previous comment. Thus, to clarify: testing whether the cache key is changed at all (contains the domain) should be in an xpcshell test if we write a separate test for it. I need a bit more time to think about the other things we want to test here. I don't want Mozilla complaining about choosing the wrong test suite when we are adding the tests to the patches we want to get merged.

Trac:
Status: new to assigned
Owner: tbb-team to arthuredelstein

gk - that's true, but I wanted to involve as many aspects of the page rendering as possible. The new patch is based on our getFirstPartyURI() API. Part of testing this is making sure that API also is behaving sanely for iframes, script src, link rel, etc etc.

I also just realized that we should probably also test Etag isolation as part of this. In previous versions of this cache isolation feature, it was covered. That may no longer be the case (or it may not be the case when the transition to Cache2 is complete).

Trac:
Keywords: TorBrowserTeam201411 deleted, TorBrowserTeam201412 added

Trac:
0001-Bug-13749.1-regression-tests-for-first-party-isolati.patch

First part posted for review (image and content patches still to come).

Trac:
Cc: gk, boklm to gk, boklm., mcs

Trac:
Cc: gk, boklm., mcs to gk, boklm, mcs

Trac:
Status: assigned to needs_review

Trac:
Summary: Write tests for network.thirdparty.isolate to Write tests for privacy.thirdparty.isolate

Please disregard the second patch here -- I'm developing a more comprehensive test that will supersede it. The first patch is ready for review, however.

Trac:
Username: michael
Cc: gk, boklm, mcs to gk, boklm, mcs, michael

Trac:
0001-Bug-13749.2-Regression-tests-for-first-party-isolati.patch

I've replaced the second patch with a new versions. Now both patches are ready for review.

Trac:
Keywords: TorBrowserTeam201412 deleted, TorBrowserTeam201501 added

Trac:
Keywords: TorBrowserTeam201501 deleted, TorBrowserTeam201501R added

To confirm that the second patch reports what we think it should, I first reverted the cache patch

> git revert 8425b9a358211777543303d2194b378b2765f144
[13749 e29d844] Revert "Bug 13742: Isolate cache to URL bar domain."
 1 file changed, 2 insertions(+), 16 deletions(-)

The result was all tests when the isolation is activated (including image caching) fail:

 0:29.04 TEST-UNEXPECTED-FAIL | chrome://mochitests/content/browser/netwerk/test/browser/browser_cacheFirstParty.js | Cache entries expected for iframe.html: 4
 0:29.04 TEST-UNEXPECTED-FAIL | chrome://mochitests/content/browser/netwerk/test/browser/browser_cacheFirstParty.js | Cache entries expected for link.css: 4
 0:29.04 TEST-UNEXPECTED-FAIL | chrome://mochitests/content/browser/netwerk/test/browser/browser_cacheFirstParty.js | Cache entries expected for script.js: 4
 0:29.04 TEST-UNEXPECTED-FAIL | chrome://mochitests/content/browser/netwerk/test/browser/browser_cacheFirstParty.js | Cache entries expected for img.png: 4
 0:29.04 TEST-UNEXPECTED-FAIL | chrome://mochitests/content/browser/netwerk/test/browser/browser_cacheFirstParty.js | Cache entries expected for object.png: 4
 0:29.04 TEST-UNEXPECTED-FAIL | chrome://mochitests/content/browser/netwerk/test/browser/browser_cacheFirstParty.js | Cache entries expected for embed.png: 4
 0:29.04 TEST-UNEXPECTED-FAIL | chrome://mochitests/content/browser/netwerk/test/browser/browser_cacheFirstParty.js | Cache entries expected for xhr.html: 4
 0:29.04 TEST-UNEXPECTED-FAIL | chrome://mochitests/content/browser/netwerk/test/browser/browser_cacheFirstParty.js | Cache entries expected for audio.ogg: 4
 0:29.04 TEST-UNEXPECTED-FAIL | chrome://mochitests/content/browser/netwerk/test/browser/browser_cacheFirstParty.js | Cache entries expected for video.ogv: 4
 0:29.04 TEST-UNEXPECTED-FAIL | chrome://mochitests/content/browser/netwerk/test/browser/browser_cacheFirstParty.js | Cache entries expected for track.vtt: 4

then I brought back that patch and instead reverted the image cache isolation patch:

> git revert 114cd22282f8b3cd6e6a5c29de8a8c396a79acc0
[13749 0236e36] Revert "Bug #6539: Isolate the Image Cache per url bar domain."
 12 files changed, 156 insertions(+), 274 deletions(-)

and the result was that image caching tests fail:

 0:32.53 TEST-UNEXPECTED-FAIL | chrome://mochitests/content/browser/netwerk/test/browser/browser_cacheFirstParty.js | Cache entries expected for img.png: 4
 0:32.53 TEST-UNEXPECTED-FAIL | chrome://mochitests/content/browser/netwerk/test/browser/browser_cacheFirstParty.js | Cache entries expected for object.png: 4
 0:32.53 TEST-UNEXPECTED-FAIL | chrome://mochitests/content/browser/netwerk/test/browser/browser_cacheFirstParty.js | Cache entries expected for embed.png: 4
 0:32.53 TEST-UNEXPECTED-FAIL | chrome://mochitests/content/browser/netwerk/test/browser/browser_cacheFirstParty.js | Cache entries expected for img.png: 1
 0:32.53 TEST-UNEXPECTED-FAIL | chrome://mochitests/content/browser/netwerk/test/browser/browser_cacheFirstParty.js | Cache entries expected for object.png: 1
 0:32.53 TEST-UNEXPECTED-FAIL | chrome://mochitests/content/browser/netwerk/test/browser/browser_cacheFirstParty.js | Cache entries expected for embed.png: 1

So I feel pretty comfortable that these regression tests are testing what they are supposed to be testing.