1420475 - Intermittent text/auto-hyphenation-ca-1.html == text/auto-hyphenation-ca-1-ref.html | image comparison, max difference: 255, number of differing pixels: 2116

Status: RESOLVED DUPLICATE of bug 1243108

(Core :: Security: Process Sandboxing, defect, P5)

Description

[Treeherder Bug Filer]

Filed by: philringnalda [at] gmail.com

https://treeherder.mozilla.org/logviewer.html#?job_id=147433480&repo=autoland

https://queue.taskcluster.net/v1/task/eAyBi_BGR9qFCKeR4YkP0Q/runs/0/artifacts/public/logs/live_backing.log

https://hg.mozilla.org/mozilla-central/raw-file/tip/layout/tools/reftest/reftest-analyzer.xhtml#logurl=https://queue.taskcluster.net/v1/task/eAyBi_BGR9qFCKeR4YkP0Q/runs/0/artifacts/public/logs/live_backing.log&only_show_unexpected=1

Comment 1

[Jonathan Kew [:jfkthame]]

The screenshot in reftest-analyzer looks like the reference file failed to load at all, and the canvas still contains content from an unrelated earlier testcase.

In the log, there's a suspicious-looking "Sandbox: Unexpected EOF" message that occurs at exactly this point. Did the sandbox mysteriously fail to let us load the reference file properly?

[task 2017-11-24T14:30:44.957Z] 14:30:44     INFO - REFTEST TEST-LOAD | file:///builds/worker/workspace/build/tests/reftest/tests/layout/reftests/text/auto-hyphenation-ca-1-ref.html | 595 / 1964 (30%)
[task 2017-11-24T14:30:44.985Z] 14:30:44     INFO - Sandbox: Unexpected EOF, op 0 flags 0302 path /dev/shm/org.chromium.q8rSGl

Comment 2

[Jonathan Kew [:jfkthame]]

Actually, I guess it's not that we couldn't load the testcase; rather, the sandbox message seems to be related to shared-memory (/dev/shm) used somewhere in graphics, in view of what comes next:

[task 2017-11-24T14:30:44.985Z] 14:30:44     INFO - Sandbox: Unexpected EOF, op 0 flags 0302 path /dev/shm/org.chromium.q8rSGl
[task 2017-11-24T14:30:44.985Z] 14:30:44     INFO - [GFX1-]: Failed to lock new back buffer.

And the graphics-related failure is probably why the canvas shown by reftest-analyzer has stale content.

So I don't know whether this failure should be blamed on Sandbox or Gfx. Either way, it's not really this testcase at fault, or Layout at all; it's just the random victim of failure at a different level.

Comment 3

[Intermittent Failures Robot]

1 failures in 744 pushes (0.001 failures/push) were associated with this bug in the last 7 days.    

Repository breakdown:
* autoland: 1

Platform breakdown:
* linux32: 1

For more details, see:
https://brasstacks.mozilla.com/orangefactor/?display=Bug&bugid=1420475&startday=2017-11-20&endday=2017-11-26&tree=all

Comment 4

[Jed Davis [:jld] ⟨⏰|UTC-7⟩ ⟦he/him⟧]

(In reply to Jonathan Kew (:jfkthame) from comment #1)
> In the log, there's a suspicious-looking "Sandbox: Unexpected EOF" message
> that occurs at exactly this point. Did the sandbox mysteriously fail to let
> us load the reference file properly?
> 
> [task 2017-11-24T14:30:44.957Z] 14:30:44     INFO - REFTEST TEST-LOAD |
> file:///builds/worker/workspace/build/tests/reftest/tests/layout/reftests/
> text/auto-hyphenation-ca-1-ref.html | 595 / 1964 (30%)
> [task 2017-11-24T14:30:44.985Z] 14:30:44     INFO - Sandbox: Unexpected EOF,
> op 0 flags 0302 path /dev/shm/org.chromium.q8rSGl

That error message means that the broker closed the request's response socket without sending a response.  The interesting thing is that there's no accompanying message from the broker (cf., for example, bug 1401774).  This is in the middle of a test run, so it's not because of racing with parent process shutdown.  So the only case that's left is where sendmsg()ing the response fails (and, yes, EINTR is handled).

I could try landing a patch to log the error code in that case, but that might not narrow it down much.

Comment 5

[Firefox Bug Husbandry Bot]

https://wiki.mozilla.org/Bug_Triage#Intermittent_Test_Failure_Cleanup