Closed Bug 1147725 Opened 6 years ago Closed 6 years ago

Intermittent test_ocsp_fetch_method.js,test_ocsp_url.js | xpcshell return code: 0 | - -8071 == 0

Categories

(Core :: Security: PSM, defect)

ARM
Gonk (Firefox OS)
defect
Not set
normal

Tracking

()

RESOLVED FIXED
mozilla40
Tracking Status
firefox38 --- unaffected
firefox39 --- unaffected
firefox40 --- fixed
firefox-esr31 --- unaffected
b2g-v2.2 --- unaffected
b2g-master --- fixed

People

(Reporter: RyanVM, Assigned: Cykesiopka)

References

Details

(Keywords: intermittent-failure)

Attachments

(2 files)

Attached file test log
See the attached log and comment 1 for more details.
Summary: Intermittent test_ocsp_url.js | xpcshell return code: 0 | - -8071 == 0 → Intermittent test_ocsp_fetch_method.js,test_ocsp_url.js | xpcshell return code: 0 | - -8071 == 0
Component: Security → Security: PSM
-8071 is SEC_ERROR_OCSP_SERVER_ERROR, which might be the result of an OCSP timeout. This seems consistent with my theory in Bug 1147726 comment 4, but again, I've only done some preliminary investigation.
See Also: → 1147726
Looks like the same deal as Bug 1147726, sadly.
Assignee: nobody → cykesiopka.bmo
Status: NEW → ASSIGNED
Attachment #8588756 - Flags: review?(dkeeler)
Comment on attachment 8588756 [details] [diff] [review]
bug1147725_disable-test_ocsp_fetch_method.js-test_ocsp_url.js-b2g-debug_v1.patch

Review of attachment 8588756 [details] [diff] [review]:
-----------------------------------------------------------------

This is ok for now, but I may have figured out the real solution to this. If I understand what we're doing correctly, we're verifying a certificate on the main thread, which involves waiting on network i/o. However, that network i/o depends on some javascript that also runs on the main thread. The current "solution" is to spin the main thread event loop: https://hg.mozilla.org/mozilla-central/annotate/883e17fc475f/security/manager/ssl/src/nsNSSCallbacks.cpp#l453 . This is a bad idea for many reasons, but in particular this may be causing our timeouts. We would need something like bug 1023621 (which I've actually been working a bit on - I'll maybe clean that patch up and ask for review later this week).
Attachment #8588756 - Flags: review?(dkeeler) → review+
(In reply to David Keeler [:keeler] (use needinfo?) from comment #41)
> This is ok for now, but I may have figured out the real solution to this. If
> I understand what we're doing correctly, we're verifying a certificate on
> the main thread, which involves waiting on network i/o. However, that
> network i/o depends on some javascript that also runs on the main thread.
> The current "solution" is to spin the main thread event loop:
> https://hg.mozilla.org/mozilla-central/annotate/883e17fc475f/security/
> manager/ssl/src/nsNSSCallbacks.cpp#l453 . This is a bad idea for many
> reasons, but in particular this may be causing our timeouts. We would need
> something like bug 1023621 (which I've actually been working a bit on - I'll
> maybe clean that patch up and ask for review later this week).

Ah, I see, makes sense.
Thanks for the review.

(The patch just disables the tests, so no try push, but I can still do one if needed).
Keywords: checkin-needed
https://hg.mozilla.org/mozilla-central/rev/e849aa3f2ed8
Status: ASSIGNED → RESOLVED
Closed: 6 years ago
Flags: in-testsuite-
Resolution: --- → FIXED
Target Milestone: --- → mozilla40
You need to log in before you can comment on or make changes to this bug.