Closed Bug 1199907 Opened 9 years ago Closed 9 years ago

TEST-UNEXPECTED-FAIL | all-test-dirs.list:toolkit/mozapps/extensions/test/xpcshell/test_hotfix_cert.js | xpcshell return code: 0

Categories

(Toolkit :: Add-ons Manager, defect)

Product:
Toolkit
Component:
Add-ons Manager
Version:
45 Branch
Type:
defect
Priority:
Not set
Severity:
normal

Tracking

()

Status:
RESOLVED FIXED

People

(Reporter: Fallen, Assigned: mossop)

References

Details

(Keywords: intermittent-failure, regression)

Attachments

(1 file)

bug1199907_hotfix_cert_test.patch
1.17 KB, patch
mossop
: review-
Details | Diff | Splinter Review 
This has been occurring on all comm trees lately. I first thought it might be fallout from addons signing, but unfortunately it is the dreaded nsIHttpServer not working after a request or two which I have also been getting in bug 1190012.

What happens is that the first request succeeds, but all other requests just time out. As a result, the addon download never finishes and the whole test times out.

http://mxr.mozilla.org/comm-central/source/mozilla/toolkit/mozapps/extensions/test/xpcshell/test_hotfix_cert.js

I've tested this with the latest comm-central clone, where I then went into test_hotfix_cert, commented out all tasks but the first and made an exact copy of the first task. This makes the test timeout as well.

I then copied the httpserver initialization into the beginning of tryInstallHotfix, so that on each task a new httpserver is started and stopped. This makes the modified test work again.

The bug must therefore be some interaction between nsIHttpServer, comm-central, and xpcshell. I spent a few days trying to debug this in bug 1190012 but I really have no idea how to fix it.

Patrick, as you do reviews for Core/Networking, would you be able to help out in trying to figure out the problem here? If you have some ideas on what debugging I can enable or what the problem can be, please let me know. Or maybe you can defer the needinfo to someone else.
Flags: needinfo?(mcmanus)
Flags: needinfo?(mcmanus) → needinfo?(jduell.mcbugs)
Fun fact: if you run this test on an IB build, it fails immediately in the second test (no timeout).

Whether that means it's more or less broken there, I don't know.
For the record:

http://hg.mozilla.org/comm-central/pushloghtml?fromchange=dd26276d7402&tochange=bdb1708f8b8c
http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=a69094e0f2a4&tochange=e537a1ba501b

WORKS:
c-c: dd26276d7402
m-c: a69094e0f2a4

FAILS:
c-c: bdb1708f8b8c
m-c: e537a1ba501b


The c-c changesets are close together and pretty unrelated, there are a lot of m-c changesets. I don't even know if it is pure coincidence. Maybe someone with a fast machine could build with these changesets, make sure they correctly work/fail, and then bisect the m-c changesets.
Keywords: regression
Might this be related to the cache patch in bug 11706464 (now backed out)? Let's see if it works now.
(In reply to aleth [:aleth] from comment #32)
> Might this be related to the cache patch in bug 11706464 (now backed out)?
> Let's see if it works now.

Nope, still failing (at least locally).
 0:02.58 PROCESS_OUTPUT: Thread-1 (pid:29609) "1447099520991	addons.productaddons	WARN	Failed downloading XM, status: 0, reason: error"
 0:02.58 PROCESS_OUTPUT: Thread-1 (pid:29609) "1447099520995	addons.manager	WARN	Failed to update system addons: Error: Failed downloading XML from http://%(server)s/dummy-system-addons.xml
....


...

So the url isn't expanded to a real server. Do we use
http://mxr.mozilla.org/comm-central/source/mozilla/testing/talos/talos/config.py#144
The above might be unrelated, or at least irrelevant.

For the failure we have now on trunk, it's a matter if singing. If i build with add-on signing enabled it works, if I build without it the test fails.
Depends on: 1168571
Disable the test for when addon-signing is disabled. 

This is a bit concerning though. I don't really understand this enough to say if it's just the test that's wrong or if hotfixinig is just not working when addon signing is disabled. If it's not just a test failure, ff dev edition (and thunderbird) can't be hotfixed.

xref bug 1210995
Assignee: nobody → mkmelin+mozilla
Status: NEW → ASSIGNED
Attachment #8687713 - Flags: review?(dtownsend)
Comment on attachment 8687713 [details] [diff] [review]
bug1199907_hotfix_cert_test.patch

Review of attachment 8687713 [details] [diff] [review]:
-----------------------------------------------------------------

Hotfixing should still be working. Hotfixes must be signed by something chaining to the AMO root if ADDON_SIGNING is enabled (dev edition) or my any cert that matches the prefs otherwise (Thunderbird). That said I'd rather we fix verifyZipSignedState to always validate the certificate for hotfix add-ons, similar to bug 1199907. Again I can do that if you want, just assign it to me.
Attachment #8687713 - Flags: review?(dtownsend) → review-
Assignee: mkmelin+mozilla → dtownsend
Component: Testing Infrastructure → Add-ons Manager
Product: Thunderbird → Toolkit
Version: 39 → 45 Branch
Depends on: 1225629
Status: ASSIGNED → RESOLVED
Closed: 9 years ago
Resolution: --- → FIXED
Flags: needinfo?(jduell.mcbugs)
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: